implanting pygmalion leadership style through workshop training

40
IMPLANTING PYGMALION LEADERSHIP STYLE THROUGH WORKSHOP TRAINING: SEVEN FIELD EXPERIMENTS Dov Eden* Dvorah Geller Abigail Gewirtz Ranit Gordon-Terner Irit Inbar Moti Liberman Yaffa Pass Iris Salomon-Segev Moriah Shalit Tel Aviv University Manager training in Pygmalion Leadership Style (PLS) was evaluated in seven field experiments. PLS is manager behavior that conveys high performance expectations to subordinates, creates a supportive climate, and attributes subordinate successes to stable, internal causes. The training workshop was developed across the seven experiments from a one-day familiarization experience to a three-day program that included learning Pygmalion concepts, skill-practice exercises, plan- ning implementation, and follow-up sessions. In all seven experiments, questionnaires measured leader and follower perceptions; in three, performance data were also analyzed. There was little evidence that the workshops influenced leaders or followers. Meta-analysis of 61 effects in the seven experiments yielded a small mean effect size (r 5 .13, p , .01). The contrast between this small effect and the medium-to-large effect produced by previous Pygmalion experiments is discussed in terms of the efficacy-effectiveness distinction. Ideas for improving attempts to get managers to be Pygmalions are discussed. * Direct all correspondence to: Dov Eden, Program in Organizational Behavior, Faculty of Management, Tel Aviv University, Tel Aviv 69978, Israel; e-mail: [email protected]. Leadership Quarterly, 11(2), 171–210. Copyright 2000 by Elsevier Science Inc. All rights of reproduction in any form reserved. ISSN: 1048-9843

Upload: independent

Post on 20-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

IMPLANTING PYGMALION LEADERSHIPSTYLE THROUGH WORKSHOP TRAINING:

SEVEN FIELD EXPERIMENTS

Dov Eden*Dvorah Geller

Abigail GewirtzRanit Gordon-Terner

Irit InbarMoti Liberman

Yaffa PassIris Salomon-Segev

Moriah ShalitTel Aviv University

Manager training in Pygmalion Leadership Style (PLS) was evaluated in seven field experiments.PLS is manager behavior that conveys high performance expectations to subordinates, createsa supportive climate, and attributes subordinate successes to stable, internal causes. The trainingworkshop was developed across the seven experiments from a one-day familiarization experienceto a three-day program that included learning Pygmalion concepts, skill-practice exercises, plan-ning implementation, and follow-up sessions. In all seven experiments, questionnaires measuredleader and follower perceptions; in three, performance data were also analyzed. There was littleevidence that the workshops influenced leaders or followers. Meta-analysis of 61 effects in theseven experiments yielded a small mean effect size (r 5 .13, p , .01). The contrast between thissmall effect and the medium-to-large effect produced by previous Pygmalion experiments isdiscussed in terms of the efficacy-effectiveness distinction. Ideas for improving attempts to getmanagers to be Pygmalions are discussed.

* Direct all correspondence to: Dov Eden, Program in Organizational Behavior, Faculty of Management,Tel Aviv University, Tel Aviv 69978, Israel; e-mail: [email protected].

Leadership Quarterly, 11(2), 171–210.Copyright 2000 by Elsevier Science Inc.All rights of reproduction in any form reserved.ISSN: 1048-9843

172 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

“We all have studies that don’t turn out.” —Gary Yukl1

The Pygmalion effect is a special case of self-fulfilling prophecy (SFP; Merton, 1948)in which raising leader expectations regarding subordinate achievement produces animprovement in performance. Rosenthal and Jacobson (1968) were the first todemonstrate the Pygmalion effect experimentally in the classroom. Although therehas been some argument that teachers’ expectations are usually realistic and thatthe Pygmalion effect is often small and practically unimportant (Jussim, 1991;Jussim & Eccles, 1995; see also Madon, Jussim, & Eccles, 1997; Smith et al., 1998),the Pygmalion effect is well established in educational psychology as confirmatorymeta-analytic results mount (Babad, 1993; Dusek, Hall, & Meyer, 1985; Harris &Rosenthal, 1985; Rosenthal, 1985, 1991a; Rosenthal & Rubin, 1978). Moreover,organizational researchers have been accumulating field experimental support forthe Pygmalion approach among adults in nonschool organizations (for reviews, seeEden, 1990a, 1993a, 1993b).

THE EARLY PYGMALION EXPERIMENTS

Phase 1 of adult Pygmalion research was launched by King (1971), who first repli-cated the effect experimentally in a nonschool organization. He created the Pygmal-ion effect among hard-core unemployed trainees in industry. Eden and Ravid (1982)and Eden and Shani (1982) increased external validity by creating Pygmalion effectsamong military personnel. Eden and Ravid also created a Galatea effect. Namedfor the statue that the mythical Pygmalion sculpted, the Galatea effect is a self-produced expectation effect among individuals who have been led to expect a lotof themselves, as distinct from Pygmalion effects that are produced by leading thesupervisor to expect a lot of the subordinates. Expanding beyond dyadic manager-subordinate relationships, King (1974) pioneered also in applying the SFP in organi-zational development (OD). He showed that raising manager expectations forthe outcomes of an OD intervention increased productivity gains. Eden (1990d)replicated the whole-group version of Pygmalion, producing the effect by raisingleader expectations toward whole groups of subordinates. Based on these earlyexperiments, theoretical explication of the Pygmalion leadership approach wasdeveloped (Eden, 1988b) and preliminary, practical implications of the Pygmalioneffect for OD (Eden, 1986, 1988a) and for consulting (Eden, 1990a) have beenexplored.

ELABORATION EXPERIMENTS

The next several experiments extended the Pygmalion paradigm to different levelsof the independent variable, to additional dependent variables, and to differentpopulations. Oz and Eden (1994) moved from the study of the effects of high leaderexpectations to a replication of the SFP effect at the low end of the expectancyscale. Babad, Inbar, and Rosenthal (1982) had coined the term “Golem effect” torefer to the negative impact of low teacher expectations on performance. However,

Implanting Pygmalion 173

for ethical reasons, Babad et al. (1982) rightly refrained from creating low expecta-tions to study their debilitating effects and instead employed a cross-sectionaldesign. Oz and Eden (1994) studied the Golem effect experimentally in an ethicallyacceptable way by averting it. By leading supervisors to reinterpret low scores inprior ability tests, they showed experimentally that uprooting naturally formed lowexpectations attenuated the Golem effect.

All the above experiments involved performance as the dependent variable.Eden and Kinnar (1991) showed that the SFP approach could be applied to raisethe volunteer rate for service in elite combat units, and Eden and Aviram (1993)showed that it could help speed reemployment among persons who had sufferedrecent job loss. Next, Eden and Zuk (1995) extended the SFP approach to sea-sickness and performance at sea. They showed that naval cadets who were led tobelieve, supposedly on the basis of psychological tests and evaluations, that theycould overcome seasickness and perform well despite rough seas reported lessseasickness and performed better at sea than their control comrades. The authorsconcluded optimistically that Pygmalion applications were limited only by our finiteimagination.

All the previously described replications involved men. Several studies werededicated to extending generalizability to women. In a two-experiment report, Dvir,Eden, and Banjo (1995) showed that the Pygmalion effect can be produced amongwomen. In Dvir’s all-women experiment, the Pygmalion effect did not emerge, andin Banjo’s experiment it emerged among men led by a man and among women ledby a man, but not among women led by a woman. Davidson and Eden (in press)then replicated the Golem findings twice among disadvantaged women led bywomen, demonstrating the remedial value of the SFP effect. Mediation analysis inthese last two experiments revealed that leader support, self-efficacy, and motivationmediated the effect; leaders whose expectations were raised were rated as betterand more supportive leaders than the control leaders, and their subordinates re-ported greater self-efficacy and motivation.

Finally, in a university-based outreach project, Natanovich and Eden (1999)replicated the Pygmalion effect among male and female supervisors of male andfemale tutors and found Pygmalion effects among all supervisor-tutor gender combi-nations. Thus, the Pygmalion effect is equally applicable to men and women. Nata-novich and Eden also produced confirmatory effects among Jewish and Arab super-visors and tutors, further extending external validity.

Thus, Pygmalion effects have been created experimentally in a wide variety ofpopulations and situations, and the role of leadership and self-efficacy in mediatingthe effects of expectations on performance has been demonstrated. These experi-mental findings validate the Pygmalion-at-work model, according to which: HighLeader Expectations → Improved Leadership → Augmented Self-Efficacy → GreaterMotivation → Intensification of Effort → Better Performance (Eden, 1990c).

OTHER PYGMALION RESEARCH

Not all adult Pygmalion research has been conducted in work or military organiza-tions. Learman, Avorn, Everitt, and Rosenthal (1990) studied SFP effects among

174 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

caregivers and their patients in a nursing home, and Chapman and McCauley (1993)found SFP effects among National Science Foundation awardees in a comparisonof seven-year follow-up Ph.D. completion rates among awardees and nonawardees.They interpreted the higher completion rate among awardees as both a Pygmalioneffect and a Galatea effect arising from high mentor expectations and self-expecta-tions aroused by being bestowed this coveted award. Chapman and McCauley thusadded higher education as a setting in which SFP effects have been demonstrated.

In meta-analyses of Pygmalion at work, Kierein and Gold (in press) found anoverall effect size of d 5 0.81 in nine studies; using different inclusion criteria,McNatt (2000) calculated an overall effect of d 5 1.13 in 17 studies. Thus, Phase1 of Pygmalion research came to its end with solid and consistent evidence for itsinternal and external validity.

PYGMALION AND DECEPTION

Nearly all published Pygmalion research employed deceptive experimental treat-ments to raise expectations. The leaders were typically duped into expecting moreand, consequently, unwittingly provided better leadership to subordinates expectedto achieve more. This means that it was nonconscious mental processes that ledthem to treat their subordinates in accordance with their expectations. Once havinginternalized high expectations, the managers produced the SFP “automatically.”Bargh and Chartrand (1999) explained how the SFP can operate “entirely noncon-sciously” (p. 467) and summarized experimental evidence for the automatic activa-tion of stereotypes in which participants nonconsciously act in accordance withtheir expectations toward others and thereby get others to act as expected with noawareness of the process by either party. The so-called “dual-process models”in social psychology (Chaiken & Trope, 1999) acknowledge both conscious andnonconscious determinants of behavior.

Thus, in principle, individuals can be aware or unaware of their role in aninterpersonal SFP process. However, the deceptive induction of high expectationsin all the Phase 1 SFP experiments renders it uncertain whether similar productiveoutcomes will be obtained when managers are conscious of the process. Perhapsautomaticity is integral to the SFP process, and inviting leaders to collaborate in afully-informed, joint attempt to produce it makes them conscious of the processand prevents it.

Moreover, giving random information to leaders in organizational settings maynot be the ideal way to build trust between the academic and practitioner communi-ties and to create lasting, positive effects. To advance the Pygmalion concept beyondbasic research to application in management as a productive approach to leadership,we must devise ways to implement it without deceiving the intended beneficiaries.Here, practical and ethical considerations converge, for basing high expectations onrandom information both raises ethical ambiguities and risks long-term consultantcredibility. The training approach used in the next generation of studies involvedthe leaders as partners in the research and obviated the need for deception.

Implanting Pygmalion 175

PREVIOUS PYGMALION TRAINING INTERVENTIONS

The only previous nonschool research on creating Pygmalion effects through train-ing was Crawford, Thomas, and Fink’s (1980) “Pygmalion at Sea” project in theU.S. Navy. Performance ratings of chronically low-performing sailors were improvedby raising expectations of both the sailors and their supervisors. This was donethrough workshops aimed at raising commanders’ negative expectations toward thesailors, counseling and guidance training to senior enlisted supervisors, and personalgrowth workshops for the low performers themselves. Comparisons of the experi-mental low performers with their shipmates and with similar low performers onother ships revealed improvements in performance and discipline among the former.There were several serious, uncontrolled threats to internal validity in this study(see Eden, 1988b, 1990c). Nevertheless, Crawford and coworkers pioneered thepractical application of the Pygmalion concept. Their intervention was designed tomake the personnel involved—both the commanders and the enlisted men—fullyinformed collaborators in creating positive SFP. However, replication to test theeffectiveness of Pygmalion training with adequate internal validity was still needed.

Weinstein et al. (1991) used a quasi-experimental design to assess a collaborative,community-oriented, preventative intervention for ninth-graders at risk for schoolfailure. The Pygmalion concept was the basis of the program. Teachers and otherschool staff, who served as Weinstein’s collaborators, communicated positive expec-tations to low achievers. They also changed such aspects of school functioning asstudent responsibility, curriculum, and evaluation in accord with the high-expec-tancy culture they were trying to create. The authors dubbed their results “promisingbut not uniform” (p. 333). Pupils who were “at risk” attained higher grades, fewerdisciplinary referrals, and better retention at year’s end. However, these improve-ments were not maintained the following year with teachers not involved in theproject. Weinstein and coworkers suggested that their intervention may not havebeen strong enough to effect lasting change.

PHASE 2—THE PRESENT APPROACH

Progress in Pygmalion research, as well as the pioneering attempts by Crawford etal. (1980) and by Weinstein et al. (1991) to apply the Pygmalion concept, sparkedreviews (Eden, 1990b, 1993a, 1993b) that were quite sanguine about the potentialfor Pygmalion applications to improve leadership and boost productivity appreciablywith relatively little investment. Buoyed by the optimism born of the success ofthe basic experiments in Phase 1, we embarked on Phase 2. This was a new,programmatic effort to develop a Pygmalion leadership workshop, to train samplesof managers in the workshop, and to evaluate the effectiveness of the training usingrigorous field-experimental design.

Pygmalion Leadership Style

Pygmalion Leadership Style (PLS) is defined as a set of behaviors that managersuse when they have high performance expectations. Based on empirical findingsand on several theoretical models (e.g., Eden, 1988b, 1990c; Rosenthal, 1991a;

176 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Rosenthal & Rubin, 1978), it is hypothesized that these leader behaviors conveyhigh expectations to subordinates and arouse high motivation and intensificationof effort. PLS includes leader behavior that creates a supportive interpersonalclimate, attributes subordinate success to stable, internal causes and their failuresto ephemeral and external causes, and motivates subordinates by strengtheningtheir self-efficacy (see elaboration in Eden, 1993b). We set out to demonstrateexperimentally that PLS training improves managerial effectiveness.

Collaborators, Not Dupes

The key difference between the seven experiments in Phase 2 and those thatpreceded them was that, whereas in Phase 1 the leaders typically had been dupedinto expecting more of their subordinates, in Phase 2 the managers were fullyinformed of the aims of the workshops at the outset, and they knew the true natureof the research in which they were participating. Thus, the new experiments testedthe applicability of the Pygmalion paradigm more than the preceding experimentshad. The managers who had produced the Pygmalion effects in the Phase 1 experi-ments did so unwittingly. However, for training to get participating managers tochoose to adopt PLS, they must be convinced of its superiority; they must becomeeager collaborators.

PYGMALION LEADERSHIP WORKSHOP

The Pygmalion leadership workshop developed into a training program that hadthree primary aims. The first was to increase participating managers’ attributedefficacy, that is, to get them to believe that their subordinates are indeed capableof achieving more. Second, to increase participants’ managerial self-efficacy bygetting them to believe that they are capable of leading their subordinates to greaterachievement. Third, several sessions were devoted to behavioral skill acquisitionto raise participants’ proficiency in enacting PLS.

The workshop opened with a 20-minute experiential warm-up exercise to demon-strate the meaning and strength of the leader-follower Pygmalion effect (see descrip-tion in Eden (1990c, p. 163). This was followed by an interactive lecture on SFPin economics, medicine, science, education, and management; Pygmalion, Galatea,and Golem effects; and leadership as the prime mediator of the effects of expecta-tions on motivation and performance. The remaining two and a half days wereinterspersed with brief lecturettes on positive and negative SFP; general and specificself-efficacy, managerial self-efficacy, attributed efficacy, collective efficacy, andmeans efficacy (Eden, 1996, 1999); the effective management of attributions in thewake of success and failure; managing organizational culture to enhance organiza-tionwide SFP. Analysis of famous Pygmalion-like leaders (e.g., Lee Iacocca) wassupplemented by examples offered up by the participants.

Next, a session was devoted to identification of the myriad opportunities toexercise Pygmalion leadership. This set the stage for role playing to rehearse self-efficacy-enhancing leader behaviors in a variety of manager-subordinate situationsusing VTR for instant replay and feedback. No structured role plays were provided;

Implanting Pygmalion 177

rather, the participants were encouraged to bring forth relevant situations fromtheir own experience. The procedure used for conducting these sessions manifestsPygmalion leadership principles: a manager describes a leader-follower situation,articulates a Pygmalion goal for the interaction, role-plays the situation, and getsfeedback, usually quite negative, from fellow participants. After viewing the videoreplay, the participant defines an improvement goal, plays the role again, and getsfeedback, almost always positive, based on improved performance of the Pygmalionrole with practice.

At the end of the role-playing sessions, the facilitator emphasized to the partici-pants how the role-playing exercise epitomized the Pygmalion leadership process by(1) encouraging them to try something difficult, (2) providing constructive feedbackabout their performance, (3) getting them to try again, leading to improved perfor-mance, and (4) providing feedback attributing the success of their ability to expandtheir mastery of PLS, thereby reinforcing their learning, enhancing their managerialself-efficacy, and augmenting their motivation to apply PLS.

The workshop concluded with implementation planning at the individual, depart-mental, and organizational levels using printed forms that lead the participantsthrough a process of defining Pygmalion leadership goals at different levels and fordifferent time spans (immediate, several months, and one year). They were askedto foresee obstacles to transfer and implementation (e.g., overload, crises thatdemand immediate attention, subordinate resistance) and to plan how to overcomethem. The workshop closed optimistically with a strong note of high expectations.

In the first several experiments described below, the workshop lasted only oneor two days. The training model developed over time and was delivered as describedpreviously in the last several experiments. In Experiment 7, we ended with asession in which bank branch managers presented their Pygmalion plans to regionalmanagement. The description of each experiment states which parts of the workshopwere included.

THE SEVEN TRAINING EXPERIMENTS

Phase 2 has, to date, included seven studies based on variations on this workshop.This is the first published report on any of these studies. Each was a master’s thesisconducted by a student under my supervision, usually with my active involvementalso in the training. In each experiment, we measured leadership as a manipulationcheck to validate the workshop as a means for getting the managers to changetheir leadership behavior. The general hypothesis tested was that PLS augmentssubordinates’ self-efficacy, motivates intensification of effort, and culminates inimproved performance. The specific variables included varied somewhat from studyto study. Table 1 shows who conducted each experiment, in what kind of organiza-tion, how much training was involved, which variables the training improved andwhich remained unchanged. In all seven, we measured the perceptions of the leadersand of the followers; in three, we also analyzed objective performance data. Sixwere true experiments in which we assigned participants randomly to conditions;one was a quasi-experiment.

Although the results were mixed, most of the experiments detected little evidence

178 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Table 1. The Seven Experiments

Student Improved Unchanged

Moti Liberman None Leadership, motivation,performance

Abigail Gewirtz Male counselors’ Female counselors’ leadership,leadership, teamwork, Jewish identity, funcommitment to youthmovement

Dvorah Geller None Managers’ GSE, Pygmalionexpectations and attitudes(MSE and PAI combined);workers’ GSE, expectations andmotivation, crew efficiency,absenteeism

Yaffa Pas Principals’ GSE, PAI Principals’ leadership, self-as-Pygmalion; Teachers’ GSE,job satisfaction

Iris Salomon-Segev Managers’ GSE and PAI Managers’ leadership and MSE;Employees’ GSE and jobsatisfaction; branch businessperformance

Ranit Gordon- None Managers’ leadership, PAI, andTerner MSE; workers’ GSE and

Pygmalion ClimateIrit Inbar None Managers’ leadership, GSE, MSE,

and PAI; employees’ GSE,domain self-efficacy, collectiveefficacy, and Pygmalion climate;branch performance

Note: GSE 5 general self-efficacy; MSE 5 managerial self-efficacy; PAI 5 Pygmalion attitudeindex.

that the workshops improved participant leadership or that they aroused any re-sponse at all among the followers. The byline quoting Yukl at the beginning of thisarticle is a rarely admitted truism. However, it is also true that the general lack oftolerance for null results leads to widespread underreporting of negative findings,giving rise to the “file drawer problem” (Rosenthal, 1979). In a break with thistradition, each such experiment is described in the following sections.

Experiment 1: Pygmalion Leadership Training for Military Training Staffs

The first two PLS experiments were done to try out the basic training approachand some exercises in a workshop format that seemed appropriate for demonstratinghow the SFP concept could be applied to move leader attitudes in a positive,Pygmalion direction. Only one day was available, providing insufficient time forskill practice or planning implementation back in the organization. We decided toaccept the challenge to make the best of the limited time available.

Sample and designMoti Liberman (1991) conducted the first experiment in the Israel Defense

Forces (IDF) School for Squad Leaders. In an after-only randomized-block design

Implanting Pygmalion 179

involving five training brigades, four with three companies and one with two compa-nies, we randomly assigned six companies to the experimental condition and eightcompanies to the control condition. Blocking by brigade permitted testing thePygmalion hypothesis while controlling extraneous variance due to differencesbetween the brigades. Assigning a company to a condition meant assigning its threeor four constituent platoons with it.

The training was delivered separately to each experimental company staff, whichincluded the company commander and his lieutenant, the three or four platoonleaders, the three or four platoon sergeants, and the 12 to 16 squad leaders, totalingbetween 20 and 26 men in each company. It is customary for IDF training staffsto get several days of preparatory training before beginning a new round of acourse. The control staffs got the standard three-day pretraining preparation, whichmade no mention of Pygmalion. As part of this three-day preparation, the experi-mental staffs got the one-day Pygmalion leadership workshop.

TrainingThe trainer of each experimental company had met with his or her company

commander prior to the workshop, fully informed him of the aims, contents, andprocess of the workshop, and put the commander in control of the workshop; thetrainer assumed a consulting role. The staff trainers were Liberman and severalprofessional IDF trainers hand-picked by Liberman for their competence and devo-tion. The workshop was divided into four sessions. The first three lasted an hourand a half each, and the last one lasted two hours. The first session was a warm-up in which the company commander presented the plan for the day to his staffand paired them off to discuss peak and trough experiences in their own personalhistory as soldiers. The company commander recovered these experiences and leda discussion analyzing the role of the soldiers’ commanders in them, looking forexamples of Pygmalion effects.

In the next segment, I provided conceptual and research input in a brief lectureand discussion session. The third segment dealt with transmitting high expectationsin the wake of failure. We screened a 20-minute Israel Television Authority docu-mentary film based on the Hollywood movie entitled “Stand and Deliver” that showshow a teacher, Mr. Escalante, boosted the math proficiency of his underachievingHispanic pupils sufficiently to attain excellent scores on the Scholastic Aptitude Test.The film includes several similar local examples. The ensuing discussion focused onleading subordinates, who had failed at something, to ultimate success. The closingsession stressed conveying positive expectations.

Three weeks after the workshop, each trainer met with the company commanderwith whom he or she had worked for a one-on-one follow-up consultation sessionon adopting PLS and other Pygmalion applications. These sessions, designed asboosters for the workshop training, took place at the field site where the companystaff was instructing its trainees. The follow-up sessions lasted about one hour.

MeasuresNear the end of the course, trainees rated their leaders’ behavior using a 17-

item index. Eden and Shani (1982), Oz and Eden (1994), Dvir et al. (1995), andDavidson and Eden (in press) have used this leadership index in IDF Pygmalion

180 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Table 2. Trainees’ Ratings of Their Commanders’ Leadership

Experimental Control

Commander M SD n M SD n F

Squad leader 3.72 0.79 500 3.79 0.92 813 0.81Platoon sergeant 4.04 0.82 310 3.88 0.91 520 11.87*Platoon leader 4.32 0.65 322 4.18 0.78 522 19.22*Company commander 4.65 0.60 343 4.46 0.75 539 32.59*

* p , .01.

experiments. As in the past experiments, reliability was high, a 5 .91. The traineesrated their own motivation using three 5-point items asking the trainee to whatextent he wanted to continue serving in his company, to go on to a command rolein the IDF, and to become an officer, a 5 .91. Performance scores used routinelyin the course were analyzed as dependent variables.

AnalysisWe analyzed the data first at the company level and then at the individual

level. There were over 1,000 men in this experiment. However, there were only 14companies (nested within five brigades) that had been randomized into conditions.Treatment was a fixed factor. Because the brigades studied were a sample of amuch larger population of brigades, brigade was a random factor. Treatment andbrigade were crossed. Therefore, testing for treatment effects required mixed-modelanalysis of variance (ANOVA) using the mean square of the Brigade 3 Treatmentinteraction as MSerror. Unfortunately, this interaction had only four degrees of free-dom, (5 brigades 2 1) 3 (2 treatments 2 1) 5 4. Hypothesis testing with sofew degrees of freedom strongly militated against detecting significant effects. Notsurprisingly, none of these platoon-level analyses was significant.

Then, viewing Experiment 1 as a quasi experiment, we reanalyzed the data atthe individual level with over 1,000 degrees of freedom. Relaxing rigor in this waydrastically augmented statistical power; it also entailed relinquishing our claim tothis being a true experiment. Following Rosenthal and Rubin (1982), we computedr to estimate effect size and the binomial effect size display (BESD) to estimatethe success-rate equivalent of r as an expression of the practical importance of thetreatment.

ResultsTables 2 and 3 present the individual-level ANOVAs. Table 2 shows that the

experimental soldiers rated the leadership of their company commanders, platoonleaders, and platoon sergeants higher than did the men in the control companies.The only leaders whose leadership was not rated higher were those lowest in thehierarchy, the squad leaders. Converting the largest effect size in Table 2 (r 5.19) to its BESD equivalent yields a 59% likelihood of an experimental companycommander’s getting an above-median leadership rating, compared to only 40%for control company commanders; the smallest BESD is a 51% versus 49% differ-ence for the squad leaders. Cohen (1988) defined r 5 .10 as a small effect and r 5

Implanting Pygmalion 181

Table 3. Comparison of Experimental and Control Trainees’Motivation and Achievement

Experimental Control

Item M SD n M SD n F

Motivation 3.68 1.13 329 3.38 1.14 468 10.37*Specialty exam 65.62 15.62 395 62.06 13.33 672 26.37*Practical exam 74.87 12.49 478 71.78 12.33 629 8.13*Physical fitness 78.91 11.47 432 75.52 11.67 605 20.90*Sharpshooting 4.37 2.57 357 3.89 2.40 540 3.71

* p , .01.

.30 as medium. Thus, the practical importance of the leadership effects displayedin Table 2 ranges from less-than-medium to nil. Note that the effect gets weakerthe lower the rank of the leader.

Table 3 shows that the experimental trainees expressed significantly greatermotivation than the controls and significantly outperformed them in a specialtyexamination, in a practical examination, and in physical fitness, all assessed objec-tively. The experimental soldiers also obtained higher scores on sharpshooting, butthe difference was not significant. The BESD equivalent of the size of the effectof the treatment on the specialty exam scores (r 5 .16) is 58% for experimentaltrainees versus 42% for the controls and for sharpshooting (r 5 .06) only 53% to47%; again, the effect ranged in size from less-than-medium to less-than-small.

ConclusionsThese results appeared equivocal but encouraging. Based on the individual-level

results, we had successfully trained the experimental leaders to employ PLS, asvalidated by their subordinates’ perceptions. The successful impact of the PLSworkshop seemed to be greater the higher the level of the leader. Moreover,enacting PLS subsequently made a difference in their subordinates’ motivation andperformance. The ambiguity was due to the different results for the different levelsof analysis. The more rigorous company-level analysis would have been highlyconvincing had it been confirmatory. We thus had experimental disconfirmationand quasi-experimental confirmation. Terpstra (1981) showed that less rigorousdesigns are more likely to yield confirmatory results in OD research; but, alas, theyare deficient in internal validity and are therefore less convincing. Nevertheless, asa preliminary attempt, even getting quasi-experimental confirmatory evidence ofPygmalion leadership training effects of nontrivial magnitude generated potentmotivation to try again.

Experiment 2: Pygmalion Goes to Summer Camp

Much previous Pygmalion research among adults had been conducted in militarysettings. External validity requires replication among nonmilitary populations.Therefore, the leaders in the remaining experiments were all from various civilianorganizations.

182 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Sample and designAbigail Gewirtz (1991) gained access to three summer camps run by the Zionist

youth movements in Great Britain. The sample included all groups of campers inthese camps. Within each camp, we assigned the groups to conditions at random.Therefore, the group was the unit of analysis. The groups were nested within camps,and treatment and camp were crossed. This, too, was a randomized-block design.Blocking by camp permitted testing the effectiveness of PLS while controllingextraneous variance due to differences between camps.

The camps had ideological goals in addition to the informal educational objectivesof most summer camps. Each camp lasted two weeks. The counselors were teenag-ers, themselves former campers, who were committed to the youth movements.The three camps had 233 campers aged 11 to 15. Camps and groups of camperswithin the camps contained a roughly equal number of boys and girls. Each groupwas led by two counselors, one male and one female, with the exception of onegroup led by two female counselors. In total, there were 23 groups; 11 randomly-assigned experimental male-female counselor pairs were trained in PLS and 12pairs of control counselors received a one-day workshop in using drama techniques.Counselors in both conditions got a follow-up conversation a week later, as abooster.

TrainingThe experimental counselors got a one-day PLS workshop in London plus a 30-

minute follow-up conversation at the campsite in Wales a week later. Gewirtz andI together cotrained this workshop and conducted the follow-up conversations. Theworkshop opened with the experiential simulation described previously, followedby a quick demonstration of the self-fulfilling nature of preconceived expectationsthat people hold of other people. The ensuing group discussion emphasized howour prior expectations of others affect our perception, treatment, and evaluationof them. Next, I gave a brief lecture on PLS and screened the training film titled“Productivity and the Self-Fulfilling Prophecy: The Pygmalion Effect.” The morn-ing ended with presentation of research findings and discussion of their practicalimplications.

The afternoon was spent “brainstorming.” The participants were split into smallgroups and encouraged to suggest ideas how to apply the Pygmalion approach toleading their campers. We summarized these ideas and concluded with several roleplays modeling the practical application of PLS concepts to camp counseling. Wegave the counselors a “Group Leader Checklist” with a page for each day of camp.We asked them to fill out on each day the way(s) in which they had applied PLSto their group that day through games, something the counselors said, or somespontaneous activity. This was intended as a daily booster to keep the counselorsmindful of the Pygmalion concepts and to prevent or reduce fade-out of the workshop.

The control counselors got a workshop of the same length on creative techniques.It was conducted by two experienced youth workers who specialized in the use ofdrama, games, and lateral thinking in presenting topics to campers. Like the PLSworkshop, the control workshop presented new concepts and provided practice

Implanting Pygmalion 183

sessions. Control counselors also got a follow-up session and filled out checklistsof their use of the new techniques they had been trained to use.

MeasuresWe adapted the Instructor Behavior Questionnaire, which Dawson, Messe, and

Phillips (1972) had used to measure college instructors’ leadership, for use withcampers. We measured teamwork using seven items selected from Eden’s (1985)18-item “Team Survey Questionnaire.” Having a positive teamwork experience isconsidered an important outcome of informal education, rather than as a meansto some other productive outcome. We measured commitment to youth movementusing a simplified, 14-item version of the 15-item Organizational CommitmentQuestionnaire (Mowday, Steers, & Porter, 1979). We measured Jewish identityusing ten questions from the 25-item index developed and modified by Miller (1986)for use with Anglo-Jewish youth. We devised five items to gauge how much thecampers achieved the camps’ secondary goal of having fun.

AnalysisJewish identity, which was measured twice, on the first day of camp and on the

last day, was subjected to a 2 (Treatment) 3 3 (Camp) 3 2 (Occasion) repeated-measures ANOVA. The other variables were measured on the last day of camponly. Each after-only dependent variable was analyzed by a 2 (Treatment) 3 3(Camp) ANOVA. As the camps studied were a sample of a larger population ofabout 15 camps, camp was a random factor. Treatment was fixed. This calls formixed-model ANOVA in which the mean square of the Camp 3 Treatment interac-tion serves as MSerror for testing the treatment effect. The Camp 3 Treatmentinteraction had only two degrees of freedom, (3 camps 2 1) 3 (2 treatments 2 1)5 2, which militated strongly against detecting statistically significant effects. Whenthe treatment effect was significant, we inspected the means to seek confirmatorypatterns of differences.

ResultsANOVA revealed significant effects of training on campers’ evaluation of male,

but not female, counselors’ leadership, F(1, 2) 5 130.00, p , .01, and on teamwork,F(1, 2) 5 53.00, p , .05. The mean ratings of the experimental male leaders exceededthose of the control male leaders in each camp as predicted. Similarly, mean team-work among groups of experimental campers exceeded that of the controls in everycamp. The positive effect of the treatment on commitment to youth movement wasof borderline significance, F(1, 2) 5 9.25, p , .10. Mean commitment was higheramong experimental groups in every camp.

Finally, there was no statistically detectable effect on achievement of the camps’prime goal of strengthening Jewish identity or on having fun. However, Jewishidentity did differ significantly between camps and across occasions. This is becausethe camps belonged to ideological movements that differed in the centrality of theirJewish values, and because the camping experience did make a difference; meanJewish identity increased between pretest and posttest in all the camps. Detectingthese significant occasion effects, which are irrelevant to our PLS hypothesis, high-

184 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

light what we did not find; the measure of Jewish identity was sensitive enough todetect explainable differences between camps and across occasions. However, thePLS training, which was evidently powerful enough to influence counselor leader-ship and camper teamwork and commitment as predicted, did not penetrate to thelevel of the campers’ Jewish values. This cannot be attributed to weakness in themeasure.

ConclusionsSuccessfully training leaders to enact PLS in a variety of practical settings seemed

more feasible after Experiment 2. The training had lasted only one day. Neverthe-less, despite extreme lack of statistical power, some of the results were significantwithout relaxing rigor. It seemed likely that lengthening the training and includingmore skill practice would strengthen the results. Thus encouraged, we went on tothe next study, this time in industry.

(Quasi) Experiment 3: Pygmalion Leadership Training for Blue-Collar Supervisors

Experiment 3 was the first that we considered to be of sufficient length to produceeffects strong enough to last and be detected in posttest measures. It was also thefirst of its kind in industry. For the sake of scientific objectivity, the roles of trainerand researcher were split; hired professionals did the training and Dvorah Geller(1993) and I did the research.

Sample and designExperiment 3 was done at the IDF Center for Overhaul and Maintenance. The

center performs periodic maintenance on armored vehicles. Though wholly ownedand directed by the IDF, the foremen and workers are civilians. In each of four ofthe center’s divisions, a roughly matched pair of production departments was chosenfor the study. One department in each division got trained and the other served asa control. The departments had an average of seven crews of ten workers each.There was a Pygmalion workshop for each experimental department head and theseven or so foremen subordinate to him. The foremen had risen to their supervisorypositions from the ranks of the entry-level workers. They were blue-collar mechanicswith no academic education and little previous management training.

TrainingEach department head and his foremen got a two-day Pygmalion leadership

workshop plus three half-day follow-up sessions; control departments got nothing.The training was done by experienced professional facilitators employed by a civiliancontractor specializing in training, OD, and managerial consulting. The trainingincluded brief lectures on Pygmalion-related topics, analysis of examples of SFP,role plays, and discussions of application.

MeasuresQuestionnaires were completed before and after the workshops. General self-

efficacy (GSE) was assessed among managers and workers by a 17-item Hebrew

Implanting Pygmalion 185

version of Sherer et al.’s (1982) GSE scale used by Eden and Aviram (1993), Edenand Kinnar (1991), and Oz and Eden (1994), pretest and posttest a 5 .87 and .90.Illustrative items are “When I make plans, I am certain I can make them work,”“When I decide to do something, I go right to work on it,” “When unexpectedproblems occur, I don’t handle them well,” and “I give up easily.”

Two 10-item, 4-point indices were devised to gauge Pygmalion expectationsand attitudes among the managers. The first, managerial self-efficacy (MSE), wasdesigned to measure the extent to which the manager felt capable of performinghis managerial role successfully. Examples of MSE items are “I can ‘get more’ out ofany worker,” “Much human potential is wasted because it isn’t managed properly,”“After a prolonged period of mediocrity it is almost impossible to improve perfor-mance,” and “Most people are as they are and there is little I, as a manager, cando to change them.” The second index was the Pygmalion Attitude Index (PAI).The PAI is a list of ten 4-point agree/disagree items that assess how efficaciousmanagers expect subordinates to be. Illustrative items are “There’s no such thingas a worker who is destined to fail,” “It is always proper to expect pleasant surprises,even from the worst workers,” “Giving lots of explanations and feedback to failingworkers is a waste of time,” and “The typical worker has already actualized hispeak potential.” A pretest among 40 students yielded acceptable reliability for bothindices, a 5 .79 and .80, respectively. However, in the study, a correlation of .67between the indices indicated lack of discriminant validity. For the sake of reliability,we merged them into one measure of Pygmalion attitudes, retaining the best 14items. Pretest and posttest reliability was acceptable, a 5 .75 and .81. Test-retestreliability, computed separately for experimental and control participants, was .46and .59, respectively.

The workers completed a 12-item index of motivation, pretest and posttest a 5.51 and .65. Also, based on conceptualization of state-trait expectancy (Eden, 1988b),we devised a 4-item measure of state expectancy to determine whether workers’performance expectations rose after their supervisor was trained. The items con-cerned the respondent’s expectations that his own performance and that of his crewwould go up or down absolutely and in comparison to other workers and crews,pretest and posttest a 5 .70 and .75.

Crew efficiency data were computed by the plant’s industrial engineers as theproduct of the standard time allotted for the production of a single unit multipliedby the number of units produced during the quarter, divided by the amount ofwork time actually invested by the crew that quarter. A high, positive score indicatedefficient use of work time. Absence was gauged by summing, for each crew eachquarter, the number of hours absent without permission and dividing it by thenumber of potential work hours that quarter. We analyzed these efficiency andabsence data for the four quarters preceding the training and for the four quartersfollowing it.

ResultsThere were no measurable treatment effects in any of the variables. None of

the pretest-posttest comparisons between experimental and control departmentswere significant, nor were any significant differences detected in the efficiency

186 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

and absenteeism data. However, in a wrap-up session with the plant manager,experimental department heads insisted that they and their departments benefittedfrom the workshops. They cited examples of how the training had improved theclimate and contributed to effectiveness. In particular, they claimed that the Pygmal-ion leadership workshop had facilitated installation of a total quality managementprogram that began shortly after the PLS workshop. After this session, plant man-agement scheduled PLS workshops for additional departments. Thus, the data toldone story, the participating managers another.

ConclusionsDespite the negative quantitative results, the managers involved felt strongly

that the Pygmalion workshops had made an important contribution. It is impossibleto know how to weigh their stated opinion. Perhaps they were being honest aboutit, and they saw real benefits from PLS training that our measures failed to detect.However, it is also possible that they insisted that the program had succeeded soas not to be associated with something that had failed. The wrap-up discussion waswith their commander, a colonel; it might be imprudent to let him believe theywere unable to make PLS work. The only way we knew to resolve such a starkcontradiction between the data and participants’ evaluations was to do anotherexperiment.

There are at least two more explanations for the disconfirmation in Experiment 3.The foremen were poorly educated and inadequately trained as managers. PerhapsPygmalion training only begins to have effects above some threshold of managerqualifications. Another obstacle was lack of enthusiasm on the part of the train-ers. Although prepared and knowledgeable in Pygmalion concepts and research,they were not completely “sold” on the model, and they included many routine,non-Pygmalion-focused “bullets” in their training. This diluted the workshop as aPygmalion treatment. These were mistakes we resolved to avoid in subsequentexperiments. One way was for me to do the training myself, even at the risk ofcompromising objectivity and incurring such criticisms as contamination, confound-ing of Pygmalion and experimenter effects, and Hawthorne effects. As it turnedout, these concerns were exaggerated because there were few effects of any kind—experimental or artifactual—in the following studies.

Experiment 4: Promoting Pygmalion to the Principal’s Office

Though the Pygmalion effect has had an illustrious past in the classroom, theeducational literature has been utterly mute about the possibility of applying theeffect in the principal-teacher relationship. Principals are managers. It seemed tous that, just as teacher expectations influence pupil achievement, and just as theexpectations of managers at large influence subordinate productivity, so principals’expectations should influence teacher accomplishment. Therefore, school principalsshould be suitable candidates for Pygmalion leadership training. This apparentlyhas never occurred to anyone in education.

By this time, self-efficacy had gained central importance in a revised Pygmalion-at-work model (Eden, 1993b). According to the model, PLS is a self-efficacy-building

Implanting Pygmalion 187

style; Pygmalion training should strengthen the self-efficacy of the participatingprincipals, and their newly acquired PLS should eventually strengthen the self-efficacy of the teachers they lead. Therefore, we included a measure of self-efficacyfor both principals and teachers in Experiment 4.

Sample and designYaffa Pass (1994) was an elementary school principal on sabbatical leave, and

an organizational behavior graduate student. She gained access to 19 elementaryschools in one district. We randomly assigned them to conditions. After one droppedout of each condition, nine experimental schools and eight control schools wereinvolved in the experiment. All 17 principals completed the principal questionnaires;135 teachers in the experimental schools and 132 teachers in the control schoolscompleted the teacher questionnaire. Both the principals and the teachers com-pleted questionnaires prior to the workshops and half a year later.

TrainingTamar Cohen and I cotrained the experimental principals in a three-day work-

shop and three half-day follow-up sessions spaced four to six weeks apart. Controlprincipals got nothing. The principals had prior knowledge about the Pygmalioneffect in the classroom. They resonated well to the notion of “principal-as-Pygmal-ion.” Training them in PLS had the flavor of training for transfer, from teacher-pupil to principal-teacher.

MeasuresThe principals and the teachers completed the GSE, a 5 .92 and .88. The

principals completed the PAI and the MSE used earlier in Experiment 3. Thepsychometric properties indicating insufficient reliability were not available whenExperiment 4 was launched. Whereas the PAI was adequately reliable, a 5 .68,the MSE was only marginally so, a 5 .56.

The teachers rated their principal’s leadership using the 17-item leadership indexused in Experiment 1, slightly modified for school principals, a 5 .94. Four 4-pointitems and a 7-point smiling-to-frowning “faces scale” were combined to measureteachers’ job satisfaction, a 5 .88.

ResultsWe tested the hypothesis using 2 (Treatment) 3 2 (Occasion) repeated-measures

ANOVA. Some significant effects were detected by the principals’ questionnaires,but not by the teachers’ questionnaires. The experimental principals acquired sig-nificantly more Pygmalion-like attitudes between the pretest and the posttest, M 53.07 and 3.37, respectively, t 5 3.75, p , .003, evidencing the predicted attitudechange in the wake of the PLS workshop. The control principals showed negligiblechange, M 5 3.41 and 3.44, t 5 0.31, not significant (n.s.) Similarly, the PLS workshopmarginally augmented the experimental principals’ GSE, M 5 3.28 and 3.46, t 51.75, p , .06, whereas GSE among the control principals remained unchanged,M 5 3.43 and 3.41, t 5 0.19, n.s.

Thus, the “Principal as Pygmalion” experiment provided some evidence in sup-

188 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

port of the effectiveness of the PLS workshop. However, this is weak evidence.Perhaps the principals, with whom we had established favorable rapport, markedthe responses they knew we hoped they would. The teachers, with whom we hadhad no contact, provided nary a morsel of confirmatory evidence. As with themanagers in Experiment 3, in a final wrap-up session, the principals respondedenthusiastically to the workshop and “wished” it upon the control principals. Theyverbalized gains in terms of their having learned how to inculcate high expectationswhile opening the new school year, how to motivate their teachers by challengingthem with a vision and strengthening their self-efficacy for actualizing it, how totreat new teachers, how to celebrate successes and manage the resulting attributionsto stable, internal factors (e.g., “You did well because you are highly capable ofsucceeding in such endeavors”), how to manage failure, and how to improve theirhandling of problem teachers.

ConclusionsThis time we had some consistent participant response and quantitative results;

both the principals’ testimony and the questionnaire data partially supported theeffectiveness of the PLS workshop, although the latter were weaker. The supportivequestionnaire data were those collected from the principals; teacher responses lentno support to the model. Nevertheless, we seemed to be getting closer to an effectiveworkshop design. We attributed this to the addition of the third day, which allowedtime for skill training, and to the openness of the principals to the Pygmalionconcept. Would it generalize to leaders less sold on the Pygmalion concept thanschool principals? The next test of the model was in a business environment.

Experiment 5: Banking on Pygmalion

A branch bank seemed to be an ideal site for an experiment. With multiplebranches it is relatively easy to randomize. A workshop for branch managers inone bank, who share many of the same problems, understand each other, and oweallegiance to the same larger organization, should be a fertile medium for the gettingacross the Pygmalion leadership message.

Sample and designIris Salomon-Segev (1994) conducted her master’s research in Bank A, a branch

bank that operates nationwide in Israel and overseas. After eliminating branchesthat were exceptionally small or large, and those in which other training or consultinginterventions were taking place, 65 domestic branches remained eligible. We ran-domly assigned 25 to the experimental condition and 40 to the control condition.Managers and all employees in the 65 branches were asked to complete question-naires before the workshops and again eight months later. However, due to restric-tions imposed by the bank on the way questionnaires were distributed and collected,response rate among employees was low. We analyzed the data provided by thebranch managers and the branch employees separately.

Due to attrition, reassignments, and failure on the part of many managers,especially in the control condition, to respond on both occasions, we could analyze

Implanting Pygmalion 189

repeated measures for only 21 experimental branch managers and 23 control branchmanagers. There were no significant pretest differences between those who com-pleted posttest questionnaires and those who did not. Including only branches inwhich at least five employees returned questionnaires on both occasions, we ana-lyzed employee data from 12 experimental and ten control branches.

TrainingTamar Cohen again joined me in cotraining. Each experimental branch manager

participated in a three-day PLS workshop. To keep the number of participants lowenough for the workshops to be effective, there were two workshops of 12 or 13managers each. The control managers received no training. The three-day formatafforded ample time for participants to practice PLS behavioral skills. For this, weconducted several hours of role playing with feedback using VCR. The three half-day follow-up sessions were used for retelling application triumphs and failures,and generating group discussion about what the participants had learned by trial-and-error in applying what they had acquired in the workshop. We analyzed failureswith the participants and together brainstormed new ways to try applying PLSprinciples.

MeasuresThe branch managers completed the PAI (pretest and posttest a 5 .51 and .65,

respectively) and the MSE (pretest and posttest a 5 .39 and .60, respectively). Dueto unreliability of the MSE, we analyzed its individual items. The employees used thesame measures as in previous studies to describe their branch managers’ leadership,pretest and posttest a 5 .96 and .95, and their job satisfaction, a 5 .85 and .86. Allcompleted the GSE, pretest and posttest a 5 .85 and .84 among the managers, and.81 and .72 among the employees.

Bank A made available performance data on the participating branches. Thesewere indicators of business success in the expert judgment of Bank A’s own experi-enced specialists. We predicted that the experimental branches would improve morethan the controls as a result of the intervention. Moriah Shalit (1996) analyzedthese performance data as part of her master’s thesis.

ResultsTable 4 presents the key results. There was no significant change in the PAI.

The strongest finding was that GSE increased among the experimental managersand remained unchanged among control managers. Furthermore, the experimentalmanagers significantly reduced their agreement with the MSE item “I set lowgoals in order not to burden workers with too many demands,” which implieslow managerial self-efficacy, whereas the control managers did not. Similarly, theexperimental managers significantly and appreciably increased their agreement withthe MSE item “Many limitations are only imaginary,” whereas control managersdecreased their agreement with it. These changes evidence a shift in the self-perceptions of the workshop participants in the direction the workshop was designedto foster. However, these are only two confirmatory items from the ten MSE items.Again, demand characteristics are a rival explanation, rendering the few significant

190 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Table 4. Comparison of Experimental and Control Branches onQuestionnaire Measures in Bank A

Item Occasion Experimental Control F (1,42)

Potential Before 3.40 3.37 0.31a

After 3.51 3.34Managers’ GSE Before 3.51 3.68 5.60*a

After 3.64 3.66“I set low goals in order not to Before 3.62 3.48 4.41*a

burden workers with too After 3.24 3.43many demands”

“Many limitations Before 2.81 3.00 4.18*a

are only imaginary” After 3.41 2.86Workers’ GSE Before 3.55 3.49 0.51b

After 3.53 3.52Workers’ perception of Before 3.91 3.59 0.30b

managers’ leadership After 3.88 3.72a n 5 21 experimental branches and 23 control branches.b n 5 12 experimental branches and 10 control branches.* p , .05.

findings among participating managers moot; they may have been trying to recom-pense us for investing our hearts and souls in their training.

The workers’ questionnaire responses evidenced no impact of the workshop ontheir perception of their managers’ leadership or on their own GSE or job satisfac-tion. Similarly, none of the business indicators detected any experimental effects.It should be noted that the business performance measures cannot be blamed.Similar measures in an experiment on a different topic in the same bank showedstatistically significant systematic variance and confirmed the experimental hypothe-sis (Eden & Moriah, 1996). The inescapable inference is that the Pygmalion traininghad no impact on the business performance of the experimental branches.

In the final follow-up session, we presented the results to the experimentalmanagers. Once again, the results aroused disbelief among the participants. Theyseemed to be much more sanguine about the benefits of PLS training than thequantitative data justified.

ConclusionsThe participants said they would have been better able to apply what they learned

had they had more support. Each had to return to his branch and face resistanceto change among department heads and clerks who had never heard of the Pygmal-ion effect. This stimulated our thinking that more hierarchical organizational sup-port for actualizing Pygmalion leadership principles was needed for application tosucceed back in the branches.

Experiment 6: Pygmalion Hospitalized

To increase hierarchical support and overcome barriers to transfer from theworkshop to the workplace, the next experiment involved training managers whose

Implanting Pygmalion 191

superiors had been trained recently. Knowing that their own supervisors had beenthrough such training should increase the participants’ openness to it. For once, asthe trainer, I could give an affirmative answer to the perennial question posed byworkshop participants: “Hey, if this stuff is so good, why haven’t our bosses gottenit?” Another advantage of training two layers of management is that the upperlevel could be supportive of the lower level in its attempts to apply the new approach.We were trying to get the managers to act in a way that would establish a Pygmalionclimate in their units.

Sample and designExperiment 6 was conducted among service managers in a large government

hospital. Services included low-level white- and blue-collar jobs performed by admis-sions and billing clerks, kitchen and dining room staffs, laundry, construction andmaintenance crews, and motor pool. In preliminary Stage 1, 16 upper-level servicemanagers participated in a three-day Pygmalion leadership workshop. In Stage 2,the experimental stage, the department heads who report to the upper-level manag-ers were randomly assigned to be trained or not to be trained. The 12 experimentaldepartment heads got a three-day Pygmalion workshop similar to the one theirsuperiors had gotten, plus one half-day follow-up session three months later; thecontrol department heads got nothing.

TrainingI conducted these workshops solo. The previous PLS training had always been

in “marathon” workshops, that is, conducted on consecutive days. The hospitalrefused to release the managers from work for a block of time. This forced aschedule of alternate days. As it turned out, rather than hindering the flow of thetraining and the intensity of the atmosphere in the workshop, the breaks appearedto fortify learning as they allowed for reflective thinking and trial-and-error betweensessions. The participants seemed to be gradually assimilating the new ideas acquiredand accumulating real-life practice using Pygmalion leadership on off-days.

MeasuresA graduate student employed at the hospital, Ranit Gordon-Terner (1994) col-

lected and analyzed the data. The managers completed the MSE and PAI, whichwere more reliable this time; pretest and posttest a 5 .76 and .72, respectively, forthe eight best items in the MSE, and pretest and posttest a 5 .64 and .63, respectively,for the nine best items in the PAI. The workers completed the GSE and theleadership index used in previous experiments. A 14-item Pygmalion SupervisoryClimate Index, especially devised for this experiment with its emphasis on climate,yielded pretest and posttest a 5 .78 and .57. Sample items asked the worker howmuch the department head “promotes the workers that are really the best ones,”“encourages the workers to give their best talents to the job,” “doesn’t really expectvery much,” and “assigns difficult tasks without regard for the workers’ ability toperform them.” Higher scores among experimental managers than among controlmanagers would indicate leader behavior that is more consistent with PLS andwould be evidence that the workshop changed their behavior as intended.

192 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

ResultsBefore-after questionnaire data among both the department heads and the ser-

vice workers yielded no confirmatory findings.

ConclusionsThis workshop was difficult to conduct. For many of the department heads, this

was their first management training. Some were quite new to their supervisorypositions, and most had risen from the ranks of the service workers. As was trueof the blue-collar foremen in Experiment 3, most of them had little formal education.Some of the workshop time had to be devoted to dealing with basic principles ofsupervision. It became apparent that not every manager can benefit from PLStraining. One must be a fairly competent manager before one can become a Pygmal-ion leader. Readiness was an important precondition that had been neglected, again,in our (my) haste to find organizations willing to participate in PLS research.

While planning this experiment and looking for a way to increase the likelihoodthat the managers would apply what they were to learn, training their superiorsfirst seemed like a great idea. Retrospectively, however, postexperimental interviewsrevealed that prior training of the upper-level managers backfired. There was nofollow-through with the upper-level managers to facilitate their application of whatthey had learned. An unexpected consequence of this design was that seeing theirsuperiors get trained in the PLS approach, but not apply it, “immunized” thedepartment heads to the workshop’s effect. Thus, although the hospital experimenthad been a move in the right direction, it seemed clear that future attempts to getmanagers to adopt the Pygmalion leadership approach would have to be based ona more complete OD model.

Experiment 7: Pygmalion-Based OD in Branch Bank B

Training the manager and his or her immediate superior was insufficient to bringabout a measurable Pygmalion effect in the hospital. The next experiment wasbased on the OD notion of intervening in the functioning of entire organizationalunits by changing their leadership and climate. The means for achieving this wasworkshop training for several managers within each of several branches of a largebank. When several managers from the same branch get trained together, theyacquire new concepts and ideas in common. They can then support each other inimplementing what they learned. This is in contrast to the branch managers in BankA, each of whom had to return to his branch alone, bringing new ideas shared byno one else in the branch.

Another innovation in Experiment 7 involved measuring several new constructs,including attributed efficacy, collective efficacy, and an expanded measure of Pyg-malion climate. Also, to increase the likelihood of implementation, the managersin each branch together planned branchwide implementation as the workshopended. Finally, they presented their plans to the regional director, who attendedthe concluding session of the workshop for this purpose. This session was designedto strengthen the hierarchical embeddedness of the PLS intervention and renderit part of the organizational reality for the experimental branches.

Implanting Pygmalion 193

Design and sampleIrit Inbar (1994) conducted her master’s research in Bank B, another bank that

has branches throughout the country. In one region of Bank B, management helpedus match pairs of intermediate-sized branches in terms of vicinity (urban neighbor-hood vs. small town), client mix (commercial vs. household accounts), businessopportunity (well-to-do vs. poor area), and ethnicity (Jewish vs. Arab). Then, werandomly assigned the branches in each pair to conditions. The limitation of 16participants in each of two workshops, or a total of 32 trained managers, and trainingfour managers from each experimental branch, dictated having eight experimentalbranches. There were thus 16 matched branches, eight experimental and eightcontrol; in addition, 259 workers in the 16 branches completed questionnaires beforethe workshops and six months later. We compared the amount of pretest-posttestchange among experimental and control branches using repeated-measuresANOVA in which each matched pair was treated as a block to control the variableson which they were matched. Therefore, degrees of freedom for hypothesis testingwas the number of paired branches minus 1, that is, 7.

TrainingTo create a Pygmalion culture by providing a shared learning experience for

several managers from the same organization, there were two three-day PLS leader-ship workshops, each for four 4-person management teams (i.e., four branches perworkshop, eight experimental branches altogether). This was followed by a half-day booster session a month and a half later plus a visit to each branch for an on-site consultation session with the participating managers. I conducted all the trainingand consulting myself. The 4-person team from each experimental branch consistedof the branch manager and the managers of the checking, credit, and foreigncurrency departments. Managers in the matched branches got nothing.

MeasuresThe managers’ questionnaire included the GSE, MSE, and PAI. Factor analysis

of the 20 MSE and PAI items led to combining the items into two indices thatyielded the highest reliability. For the four best MSE items, pretest and posttesta 5 .70 and .52; for the best eight PAI items, a 5 .72 and .57.

The workers completed the leadership (a 5 .79 and .80), GSE (a 5 .74 and .72),and motivation (a 5 .86 and .89) indices used in the previous studies. In addition,we devised three new measures for this experiment. To measure service self-efficacy,respondents rated the level of personal, professional, fast, and accurate service heor she could provide using 5-point scales ranging from “poor” to “excellent” (a 5.65 and .77).

We defined collective efficacy as the extent to which the employee perceived hisor her branch was capable of carrying out its tasks and accomplishing its goalssuccessfully. To construct the collective efficacy index, we first recast the 17 individ-ual-level GSE items into collective terms (e.g., “This branch doesn’t relent untilthe problem at hand is solved,” “This branch avoids getting involved in things thatseem too complicated”). These were 4-point agree-disagree statements. Next, wecomposed new items to cover aspects of collective efficacy not captured by the

194 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

GSE items (e.g., “This branch runs like a well-oiled machine,” “When we reallywant to, this branch can succeed at any mission,” “Most of the people in this branchare below average,” “In the face of a real challenge, this branch goes to pieces”),a 5 .92.

Finally, we gauged Pygmalion climate in terms of employee perceptions of man-ager behavior and organizational procedures that may foster or hinder the emer-gence of PLS. This scale had 32 items, half Pygmalion and half Golem, and wasdesigned for factor analysis. The two-factor solution with a varimax rotation yieldeda Pygmalion factor (a 5 .86) and a Golem factor (a 5 .81). Examples of pairs (inthe questionnaire the items were randomized, not paired) of Pygmalion and Golemitems are “In this branch usually the very best workers get promoted,” “Onlyseniority and ‘pull’ determine promotions and raises here”; “When things get tough,the people in charge inspire a feeling that the workers will come through” and“The people in charge of this branch do not believe in the workers’ capacity tocope with stress.”

Bank B made available before-after performance data on the 16 branches. Thesewere indicators of business success in the expert judgment of Bank B’s experiencedspecialists. Again, Moriah Shalit (1996) analyzed the performance data as part ofher master’s thesis.

ResultsThere were no significant effects of the intervention on any of the measured

variables. The intervention failed even to influence the managers’ leadership behav-ior measurably, let alone to affect their subordinates’ self-efficacy, collective efficacy,motivation, perceived Pygmalion climate, or branch business performance. Theseresults were not due to the lack of statistical power owing to the meager numberof degrees of freedom. The means simply did not differ; the numerators of thecritical ratios were so close to zero that even a drastic increase in degrees of freedomwould not have rendered them statistically significant.

ConclusionsOf all the seven experiments, this one was based on the most highly developed

intervention model, and it was bolstered by improved measures of the key variablestargeted for change. Still, it did not get the job done. It seems as though the “better”the design gets, the worse the results. The inescapable conclusion is that furtherincremental improvements in the design of the intervention or of the measures areunlikely to succeed. A new approach to getting Pygmalion leadership to work isneeded.

META-ANALYSIS OF THE SEVEN EXPERIMENTS

To estimate the typical effect across the seven experiments and to test its statisticalsignificance and its variability, their results were cumulated using meta-analyticprocedures outlined by Rosenthal (1991b). According to Rosenthal (1995), theresults of two or more studies can be meta-analyzed. However, “when there arevery few studies, the meta-analytic results are relatively unstable. When there arevery few studies available on a given research question, it would be more economical

Implanting Pygmalion 195

of journal space and editors’ and reviewers’ time to incorporate the meta-analysisas an extension of the last study in the series of a few studies” (p. 185). Meta-analysis of the present experiments seems especially apropos, given its power toovercome their major design weakness, which is extreme lack of statistical power.

The seven experiments yielded a total of 61 effects in a combined sample of180 organizational units (e.g., bank branches, summer camps, infantry companies;Liberman’s results, as the others, were included at the group level). Thirty-nineeffects were in the predicted direction and 17 were in the opposite direction; fivedifferences were virtually zero. Weighted by sample size, mean r was .13, F 5 2.95,p 5 .0016, one-tailed. The comparison of the magnitude of the effect sizes was alsosignificant, x2 (60) 5 136.10, p , .00001, one tail. Major contributors to overalleffect size and to effect size variability were two huge effects (r 5 .99 and .98) inthe summer-camp study that resulted from slight experimental control differences,but near-zero error terms that grossly inflated the critical ratios and effect-sizeestimates. Without these two effects, the mean of the remaining 59 effect sizes wasr 5 .108. Thus, although significant owing to the extraordinary power of meta-analysis due to the accumulation of degrees of freedom, the overall positive impactof the seven experiments was weak.

The statistical significance of the overall effect should not divert our attentionfrom its small size. The BESD equivalent of r 5 .13 is a success rate of 56% in theexperimental group compared to 43% in the control group. Few organizationswould invest scarce resources for such meager returns. A much stronger cumulativeeffect is to be expected from an intervention designed to improve leadership.Furthermore, it would be of little comfort to executives who invested dearly in aleadership development program to know that, although in their organization theeffect was not significant, it was broadly significant across a larger number oforganizations. A weak-but-significant effect size is a disheartening basis for practicalapplication.

WHY HAS IT NOT BEEN WORKING?

The reader may be tempted to discount the present disconfirmations as characteristicof masters research. However, Phase 1 included nine published, confirmatory, field-experimental replications of the SFP effect among adults conducted by my mastersstudents under my supervision (Davidson & Eden, in press [two experiments, bothconfirmatory]; Dvir, Eden, & Banjo, 1995 [two experiments, one confirmatory];Eden, 1990d; Eden & Kinnar, 1991; Eden & Ravid, 1982; Eden & Shani, 1982;Eden & Zuk, 1995; Oz & Eden, 1994). The students who conducted the Phase 2experiments were as competent as those who conducted the earlier ones. Therefore,it is unlikely that the Phase 2 studies produced such weak effects simply becausethey were masters theses. The reasons for the present disconfirmations must besought elsewhere.

Too High Expectations

We may have been expecting too much. Perhaps no three-day training workshopcan effect improvements in leadership, self-efficacy, motivation, climate, or perfor-

196 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

mance. There have been precious few highly internally valid research evaluationsof the outcomes of such training. In the first edition of Leadership in Organiza-tions,Yukl (1981) commented, “It is unfortunate that very few management develop-ment programs include rigorous procedures for evaluating training effectiveness” (p.284). Yukl’s own selective tally of “most” of the research that met his requirements atthe time for “a well designed study” (i.e., includes an untrained control group andsome external criterion of improvement) included only 26 studies over the 26-yearperiod between the publication of the first (in 1953) and last studies (in 1979). Most(but not all) of the 26 were corroborative (Yukl, 1981, Table 10-1). Yukl concludedthat training in leadership skills or managerial motivation “can be quite effectivefor improving managerial skills, altering leadership behavior, and strengtheningmanagerial motivation” (p. 284). Although Yukl discontinued the tally in the threesubsequent editions of his book, he has maintained his generally positive conclusionregarding leadership training effectiveness as confirmatory results continue to tricklein. Clearly, his conclusion that training may improve leadership is correct. However,at the rate of less than one corroborative study a year, considering the countlessuntallied studies that presumably lie in the file drawer (Rosenthal, 1979), it is alsoclear that training has not been shown to improve leadership very often.

For their meta-analytic review of managerial training research, Burke and Day(1986) found only 70 studies that met their minimal methodological requirements(i.e., included managers or supervisors, evaluated the effectiveness of a trainingprogram, and included at least one comparison group). Beyond that, “The degreeof methodological rigor or quality of the research design was not a selection crite-rion” (p. 233). Thus, the studies reviewed were of mixed quality, and none was atrue experiment; as Terpstra (1981) has shown, the least rigorous studies are theones most likely to confirm the hypothesis. Of the 70 studies meta-analyzed, Burkeand Day classified 31 as “Human Relations” studies because they were focused on“leadership, supervision, attitudes toward employees, and communication” (p. 233).Of these 31, 17 reported results in terms of “subjective behavioral criteria,” thatis, measures of change in on-the-job behavior as perceived by trainees, peers, orsupervisors. Thus, these 17 are the studies most comparable to the present sevenexperiments. Their overall average effect size was ES 5 .39, which is equivalent toan r of .19. Although somewhat larger than the average effect size in the presentexperiments, the difference is not that great, and it still falls appreciably short ofthe ES that Cohen (1988) defined as “medium” (i.e., .50).

The dearth of relevant research has been decried by many, some in quite bluntterms. Conger (1992) surveyed leadership training programs and came to a startlingconclusion, considering the multibillion dollar annual outlay for training: “In fact,no one has ever taken the time to seriously investigate what is offered in leadershiptraining programs and to ascertain whether any of them work. . . . ” Therefore,“relatively little is known about [leadership training’s] effectiveness” (p. xii). Fiedler(1996) similarly concluded that “all of the reviews of leadership training . . . stressthat we know very little about the processes of leadership and managerial trainingthat contribute to organizational performance. At least one reason for this lack ofknowledge is the scarcity of meaningful and rigorous research” (p. 244). Finally,according to House and Aditya (1997), “There is little evidence that charismatic,

Implanting Pygmalion 197

transformational, or visionary leadership does indeed transform individuals, groups,large divisions of organizations, or total organizations, despite claims that they doso . . . There is no evidence demonstrating stable and long-term effects of leaderson follower self-esteem, motives, desires, preferences, or values” (p. 443).

Thus, despite the overall upbeat tone of Yukl’s early conclusion, the evidencethat he mustered in support of this conclusion was meager, and there is no consensusamong leadership scholars that an abundance of confirmatory evidence supportsthe conclusion that training improves leadership or its outcomes. An updated meta-analysis of this literature is overdue. Meanwhile, the disconfirmations evident inthe present experiments may be characteristic of leadership training in general,not only PLS training. Viewed against the backdrop of so few previous rigorousconfirmations, perhaps the present overwhelmingly negative results are not anaberrant outcome.

Too Weak an Intervention

The training itself must have more built-in mechanisms designed to ensure post-training implementation. A two- or three-day workshop to familiarize participantswith basic concepts and to augment their rudimentary behavioral skills may be agood start. However, this should be an initial phase, to be followed by more intensiveintervention. Future applications should add more developmental activities afterthe workshop including (1) periodic meetings with a consultant who works individu-ally with each leader and supports Pygmalion implementation, (2) periodic progressreports, and (3) ongoing involvement of some higher organizational authority whooversees implementation of the Pygmalion project. In short, successful inculcationof PLS may require deeper penetrating OD-style interventions.

Consciousness Versus Automaticity

Bargh and Chartrand’s (1999) distinction between “automatic” behaviors emittedwith no awareness on the part of the individual and action taken consciously mayexplain the different results in Phases 1 and 2. The leaders who confirmed thePygmalion hypothesis in the Phase 1 experiments were volitionally neutral; theywere unaware of the true aims of the research and could not have consciouslyhelped or hindered. The manipulation checks confirmed that, if they were skepticalof the information conveyed to raise their expectations, it did not stop them fromraising their estimates of their subordinates’ potential. Having no reason to doubtwhat we told them, they believed their subordinates were qualified and “automati-cally” treated them to superior leadership. In none of the early experiments didwe say anything to get the leaders to change their leadership style; once theirexpectations were raised, they altered their behavior unwittingly, and they noncon-sciously fulfilled their experimentally induced expectations.

Conversely, the leaders in Phase 2 were fully informed of the SFP process. Nothaving been duped into expecting more, their awareness may have blocked theSFP process. Perhaps making the process conscious undermines it and renders itinoperative. If awareness blocks SFP, the training approach is doomed to failure.

198 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Widespread SFP application hinges on finding ways to get fully informed, aware,and willing leaders involved in producing Pygmalion effects.

We note that the subordinates’ responses to their leaders’ expectations are pre-sumed to be “automatic.” No assumption was made in the previous experiments,nor in the present ones, that the subordinates’ awareness plays any role in the SFPprocess to which they become a party. This may need rethinking.

Efficacy Versus Effectiveness

It is possible that, despite its considerable internal and external validity, demon-strated in Phase 1’s dozen or so overwhelmingly confirmatory experiments, thePygmalion approach to leadership lacks “application validity,” that is, the purposiveapplicability that would render it an effective tool in the hands of practicing manag-ers. This goes beyond Campbell and Stanley’s (1966) “external validity.” Seligman(1995, 1996; see also Seligman & Levant, 1998, and Holon, 1996) has used the terms“efficacy” and “effectiveness” to distinguish between recovery rates produced bypsychotherapy in controlled studies and in clinical practice, respectively. Accumu-lated findings from evaluation research show that some treatments achieve betterrecovery rates in controlled studies than in clinical practice whereas others workbetter in actual practice than in the highly controlled studies. Borrowing Seligman’sterms, although Phase 1 Pygmalion experimentation confirmed that the Pygmalionapproach has strong efficacy, the Phase 2 results show that it has not passed theeffectiveness test—yet.

A challenge for any leadership theory to is to survive the crucible of effectivenesstesting. The effectiveness record of leadership training in general seems deficient.A current exception is the field-experimental confirmation of the effectiveness oftransformational leadership training among IDF officer cadets (Dvir, Eden,Shamir, & Avolio, 1999). This came in the wake of numerous previous confirmationsof transformational leadership theory in causally ambiguous cross-sectional researchand in laboratory experiments that showed transformational leadership’s efficacy,but not its effectiveness. Dvir and coworkers have made the first stride towardestablishing the effectiveness of transformational leadership training. Like mosttheory-based approaches to leadership training, Pygmalion training still lacks sup-port for its effectiveness.

Methodological Considerations

Manipulation checkPerhaps we have not been using the right—or the only relevant—manipulation

check. The first aim of the training was to get managers to believe that theirsubordinates were capable of achieving more. The workshops were designed toraise managers’ expectations and to influence them to use PLS. Therefore, asmanipulation checks, we had a measure of leadership behavior in every experimentand a measure of Pygmalion attitudes in most of them. However, we neglected tomeasure each manager’s specific expectations toward each of his or her subordinatesas we had in all the Phase 1 experiments. Perhaps the workshops failed to raiseexpectations. If the managers’ expectations were not raised, the Pygmalion hypothe-

Implanting Pygmalion 199

sis was not tested. The PAI was intended to measure Pygmalion attitude, includingglobal expectations regarding what workers are capable of achieving. Unfortunately,it was not a highly reliable measure. It could have been at best a proxy measurefor expectations.

Thus, it is possible that these seven experiments are discontinuous with theprevious SFP research. It will be interesting to see whether future meta-analystsinclude the Phase 2 experiments as Pygmalion experiments. Thus, though it wasnot a mistake to measure leadership and Pygmalion attitudes, it was a grave oversightto neglect measuring expectations. In future replications, managers’ expectationsmust be measured.

Statistical powerAn obvious deficiency in the seven experiments was paucity of degrees of free-

dom. Lack of statistical power renders it unlikely to detect significant effects in theindividual experiments, if there were any. This austere approach to degrees offreedom may appear to exact too high a price in terms of Type II error. However,the Pygmalion experiments in Phase 1 adopted a similar approach to degrees offreedom. Despite the stringency, most of that research was confirmatory, and theeffect sizes were moderate to very large. If group-level analysis did not hamperrepeated detection of the Pygmalion effect in Phase 1, it cannot be blamed for thegeneral lack of confirmation in Phase 2.

It will be difficult to launch larger-scale experiments with appreciably moredegrees of freedom. For example, working in Bank B singlehandedly with eightexperimental branches was a formidable task that required—after planning, con-tracting, coordinating, scheduling, and preparing—conducting two three-day work-shops plus the follow-up sessions and travel to eight sites for follow-up consultations.Still, the eight matched pairs afforded only seven degrees of freedom for analysis.Even doubling the number of branches (and the training effort) would have yieldedonly 15 degrees of freedom. The significant meta-analysis is of little solace in thisregard. A way must be found to have more powerful analyses in the individualexperiments.

However, this must be done without improperly inflating the degrees of freedom,as by analyzing the data at the individual level when whole groups were assignedto conditions. The latter is a fairly common analytic procedure. For example, Barling,Weber, and Kelloway (1996) randomly assigned 20 bank branch managers to train-ing and control conditions and analyzed their data at the individual level, takingthe 80 or so subordinates instead of the 20 managers as the basis for countingdegrees of freedom. Similarly, Smoll, Smith, Barnett, and Everett (1993) trainedeight Little League coaches and took 150 degrees of freedom, based on the numberof boys coached by the eight experimental and ten control coaches, in most of theiranalyses.

Klein, Dansereau, and Hall (1994) have distinguished between the level of atheory, the level of a construct, the level of measurement, and the level of statisticalanalysis. Levels of a theory are distinguished in terms of whether its constructstreat individuals in a group as being independent, homogeneous, or heterogeneous.This hinges on the crucial issue of whether the theory’s constructs are intended to

200 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

explain variation that results from differences among individuals irrespective ofgroup membership, from differences among groups, or from differences amongindividuals within groups, respectively. Experimenters such as Barling et al. (1996),Smoll et al. (1993), and the present authors should examine differences amonggroups subordinate to leaders who were trained and control leaders; there is nothingin the leadership theories tested that hypothesizes differences among individualsirrespective of group membership or differences among individuals within groups.Individual-level analysis of such data exemplifies what Klein et al. (1994) dub“nonconformity” of levels in which the level of statistical analysis is inappropriateto the level of intervention. Thus, intervening at one level and analyzing the dataat another level—one more congenial to hypothesis confirmation—can produceresults that look confirmatory.

When we test the hypothesis at the level of statistical analysis that conforms tothe level of intervention, hypothesis confirmation is much less likely. This washighlighted by our two-level analysis of the results of Experiment 1. When managersare randomized and data are collected from their subordinates, individual-levelanalyses should be accompanied by redefinition of the study as a quasi-experiment,and the results should be discussed with extreme caution. Thus, although it isimportant to find a way to field-test leadership hypotheses with ample degrees offreedom, employing wrong-level analyses is not the way to do it.

Rigor versus organizational climateA limitation imposed by experimental rigor is that random assignment of units

at the same organizational level to experimental and control conditions rules outcreation of an overreaching supportive climate involving a larger portion of theorganization. For example, in Bank B, we could have asked the regional directorto implement Pygmalion leadership in regional meetings and in his other dealingswith the branch managers. However, had he done so with experimental and controlbranch managers present simultaneously, that would have exposed the controlmanagers to some Pygmalion leadership treatment, contaminating the experiment;treating them all to Pygmalion leadership would have been commendable manage-ment, but it would have undermined the logic of our experimental design. Similarly,he could not have discriminated between branch managers when all were presentand treated some to Pygmalion leadership and others not. Future replication mustuse a research paradigm that enables implementation in an entire section of anorganization.

MeasurementFinally, the measures designed to gauge the Pygmalion-relevant variables, PAI

and MSE, have been psychometrically borderline or worse. Although Experiments 3through 7 were described previously one after the other, in actuality they temporallyoverlapped, preventing the successive item analysis, revision, and retesting to in-crease reliability. Nevertheless, weak reliability is not the reason Phase 2 resultswere disconfirmatory. There were some reliable measures in every experiment, sothat every one of them could have provided supportive results, had the traininghad any effects. For example, the reliable leadership index that mostly failed to

Implanting Pygmalion 201

Table 5. Proposed Improvements for Future PygmalionLeadership Style Training

Enlist participants as collaborators in the researchBuild-in resistance to transfer problemsMore follow-up sessions after the training phaseMore on-site follow-up consultationsMore Pygmalion action planningMore involvement of higher managementEstablish regional Pygmalion cultureInvolve participants in devising criterion measuresGet subordinates involvedCombine Pygmalion with other approaches to leadershipSelect for readiness

detect effects in these experiments did detect effects in many Phase 1 studies.Therefore, unreliability is not the explanation for the lack of confirmation in Phase2. Nevertheless, future replications should include valid measures of the basicPygmalion variables: (1) a measure of the managers’ beliefs in the subordinates,(2) a measure of the subordinates’ perceptions that the managers act on such beliefs,and (3) measures of performance.

RECOMMENDATIONS FOR FUTURE PYGMALION APPLICATIONS

Table 5 lists the lessons learned from conducting the seven experiments. They aresummarized below in the form of recommendations to be implemented in futurePLS programs.

Enlist participants as collaborators in the research. For training to alter PLS,participants must be convinced of its worth and willing to exert effort to adopt it.More of what we know about persuasion and attitude change should be incorporatedinto the training. An effective way to win participants over “to our side” might beto invite them to join us as partners in the entire project, including the research,much like Weinstein et al.’s (1991) community approach.

Build in resistance to transfer problems. More must be done during the workshopphase to prepare the managers for coping with setbacks in their initial efforts toimplement Pygmalion leadership. There is field experimental evidence that this canbe accomplished through cognitive sensitization (Caplan, Vinokur, Price, & vanRyn, 1989). Emphasis should be placed on strengthening participants’ resilience sothat they recover quickly from the inevitable early failures encountered as theybegin implementation back on the job.

More follow-up sessions after the training phase. Participants (and consultantsand facilitators) must realize that the project is not “over” when the workshop ends,but when they adopt PLS and improve their leadership effectiveness. Participants inthe past have said almost unanimously that more follow-up sessions would behelpful to get the new approach entrenched. Two or three sessions spaced a monthor two apart have proved insufficient.

More on-site follow-up consultations. The consultant should visit participating

202 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

managers about once every two weeks for, say, the first six months after the work-shop, and then about once a month until both feel they have achieved their objec-tives. These consultations should focus on what the manager and other staffers aredoing to implement PLS. In these sessions, the consultants should model the PLSbehaviors they are encouraging the managers to adopt. The consultant should bethe same person who trained the workshop. This will lend the consulting a strongcoaching flavor. This weaves training, consulting, and coaching into an integratedsupportive fabric that casts managers in the role of Galatea whose leadership is tobe influenced by the Pygmalion consultant, ultimately developing the managersinto Pygmalion leaders.

More Pygmalion action planning. Implementation planning in previous studieswas done toward the end of the workshop, but was not formalized as a plan withquantifiable goals, deadlines, and reporting functions. In the future, special attentionshould be devoted to implementation planning in the initial workshop. Participantsmay continue working on their plans after the workshop until they are finalizedand submitted to a higher authority (e.g., the regional director) for review. He orshe will review each manager’s plans with the consultant and return them to themanager for revision. The final version of the plan should be formalized in a meetingbetween the manager who submitted it, the regional director, and the consultant.The written plan can serve as the basis for a quarterly review to be held by theregional manager with each manager on the basis of progress reports that branchmanagers will be asked to file. The written plan can also serve as a reference pointfor the posttraining consultations.

More involvement of higher management. Besides monitoring the progress of thereporting managers in implementing their PLS plans, the regional manager can beasked to draw up a PLS plan for the regional headquarters office and for the regionas a whole. He or she could also be called upon to implement PLS in interactionswith his or her staff and with all subordinate managers. This will give the regionalmanager a major role in enacting the leadership ideal that all managers in theregion are expected to adopt. The regional manager could be encouraged to presenta vision before the region, use the PLS terminology taught in the workshop, tostrive to strengthen the self-efficacy and collective efficacy of all personnel in theregion, and to implement whatever other components of PLS he or she can. Amajor role for the consultant is to coach the regional director in implementing PLSvis-a-vis the entire region.

Establish regional Pygmalion culture. The consultant should work with the re-gional manager to create changes at the cultural level. This could be accomplishedby the regional manager’s own leadership style and by installing special campaignsthat accentuate PLS concepts (e.g., Pygmalion-of-the-Month and Pygmalion-of-the-Year awards). Regional memos, minutes, notices, and greeting messages on personalcomputer screens could be spiced with Pygmalion messages.

Involve participants in devising criterion measures. Perhaps the measures in previ-ous PLS experiments have missed the essence of the change in leaders’ approachand followers’ perceptions. Future research should include subjective measures ofwhether the managers “bought” the Pygmalion approach. Also, employees should

Implanting Pygmalion 203

be interviewed to try to get data indicative of changes wrought by the interventionbut not detected by the formal measures.

Get subordinates involved. In past PLS work, the nature of the project was notdisclosed to the subordinates. We have treated them as objects through whommanagers are trying to get something done with a hidden agenda. This might be amistake. If they knew what was going on and why, employees actively involved inthe program might play a constructive Galatea role.

Combine Pygmalion with other approaches. The Pygmalion approach sharescommon ground with other approaches to leadership, such as transformational,charismatic, and visionary leadership (Avolio & Bass, 1988; Avolio & Gibbons,1988; Bass, 1985, 1998; Conger, 1989; House, 1977; House & Howell, 1992; House &Shamir, 1993; Sashkin, 1988). Future PLS applications should incorporate moretransformational, charismatic, and visionary concepts to strengthen its theoreticalbase and to enrich its potential for transfer and implementation. These other ap-proaches are compatible with Pygmalion-style leadership and would add dimensionsof application beyond those encompassed within the Pygmalion-at-work model.The resulting model contamination would be the price of finding a combinationthat works.

Select for readiness. Participant readiness is an important consideration for thesuccess of training (Harris & Fleishman, 1955). Similarly, it is crucial for the successof any leadership style to consider the situation in which it is to be implemented(e.g., Fiedler, 1967; Hersey & Blanchard, 1982; Vroom & Jago, 1988). Avolio andBass (1995) have referred to the “embeddedness” of a leadership program in anorganizational environment. The seven experiments were done with no regard forthe readiness of either the managers or the organization for assimilating the Pygmal-ion approach. Therefore, some of these applications were not adequately embeddedin the larger organizational context. Future trials should be conducted only inorganizations chosen for their appropriateness as judged by management, by inter-nal change agents or human resources specialists, and by the consultant. In particu-lar, the manager of the experimental region must be selected carefully to ensurehis or her suitability for the pivotal role of epitomizing Pygmalion leadership.

Beyond requiring an organization that is ready for change, the participants mustbe ripe. PLS training will not turn individuals who are not managers into managers,nor is it likely to turn bad managers into good managers. It is reasonable to expectsuch training to make good managers better. Some of the participants in some ofthe seven experiments were not ripe for such training. Thus, the organization andits managers should be preselected for readiness.

EPILOGUE

These proposals come in the wake of successive efforts over several years to makea practical tool out of the Pygmalion paradigm. Phase 1 had established that in itsdeceptive, five-minute version, the Pygmalion effect works. These results haveattained widespread recognition, and the Pygmalion phenomenon has been citedas kindred to the cutting-edge approaches to managerial leadership (e.g., Bass,1985; Bennis & Nanus, 1985; House, 1977; Vecchio, 1997; Yukl, 1998). However,

204 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

as Yukl pointed out in the byline, we all have studies that don’t turn out. Phase 2included seven experimental attempts to bring this effect to practical fruition that,on balance, had little impact. The disparity between the present, largely ineffective,three-day-plus-follow-up training interventions, on the one hand, and reportedlysuccessful, one-day (Barling et al., 1996) and two-and-a-half-hour (Smoll et al.,1993) interventions is irreconcilable. Future research or analysis will have to explainhow a couple of hours of training can produce measurable outcomes, whereas threesolid days of intensive training plus follow-up activities do not. This may be aninstance of Terpstra’s (1981) inverse rigor-confirmation phenomenon; the stringentrigor of these seven experiments may have obscured “effects” that less rigorousresearch may “detect.”

The contrasting results of Phases 1 and 2 show that, although the Pygmalioneffect has strong efficacy, it has only weak effectiveness. Phase 2 showed that thephenomenon shown to be externally valid in Phase 1 is not easily harnessed as amanagement tool. There is a meta-implication for applied social science in generalto be learned from this: one cannot be sure a finding is applicable until one hastried to apply it. This may be a corollary of Kurt Lewin’s famous principle that thebest way to learn about a social system is to try to change it. Conducting the researchin the field is not enough. External validity as we have known it is not enough. Aphenomenon can be supported by research that has high external validity and stilllack applicability. The implication for all leadership research is the same: untilsupported by implementation-based, field-experimental confirmation, judgment re-garding the effectiveness of any approach must be suspended. This highlights theimportance of Dvir et al.’s (1999) demonstration of the effectiveness of transforma-tional leadership training.

Another lesson to be learned from this may be that training is not a panaceafor every management problem; not everything can be trained. Perhaps the Pygmal-ion effect cannot be produced except by duping leaders into expecting more. APygmalion manager might be like a running back in football: “When you see agood one, sign him up and turn him loose on the playing field.” The conventionalwisdom among National Football League coaches is that outstanding backs areborn; they cannot be trained. Perhaps the same is true of PLS; some managers,teachers, commanders, coaches, and consultants have it naturally and some do not,and those that do not cannot be trained, coaxed, or coached to have it. Recallingthe quotation from Conger (1992), the trainability question is relevant to leadershipin general, not only to Pygmalion leadership.

In an early article (Eden, 1984), after having conducted only two Pygmalionexperiments, I posed the question whether unawareness was an indispensable ingre-dient in the Pygmalion effect. After 15 years of research on organizational SFP,this question continues to haunt me. The automaticity-versus-consciousness issueimplies an exasperating irony. Ethical and practical considerations appear to conflictafter all. If awareness prevents SFP, then by taking the ethical high ground andinforming participants of the nature of the Pygmalion process, we render it ineffec-tive. The overwhelmingly confirmatory results of the earlier experiments with unin-formed leaders, and the largely negligible results of the training experiments seem

Implanting Pygmalion 205

to imply that, in order to produce the effect, we have to compromise our ethicalpurity and dupe managers into believing in their subordinates. If managers haveto be tricked into enacting PLS, many practitioners would prefer more ethicallypalatable approaches to improving effectiveness.

Two inescapable questions arise from this dilemma. Can a manager or mentorproduce the Pygmalion effect willingly? If so, is it ethically acceptable to do so?Based on my own personal experience and values, contrary to the present experi-mental evidence, the answers to both questions are affirmative. Over the years, Ihave coached numerous students into outstanding achievement, even when I har-bored some initial doubts regarding their ability to excel. Acting Pygmalion helpspeople. Treating people as though we expect much of them does motivate them tomake the best of what they have. Ethically, it would be unacceptable to treat themin any other way. Given the positive outcomes of my personal experience consciouslyapplying PLS, I would prefer to defend the ethics of using it than to defend theethics of refraining from using it.

For me personally, the efficacy-effectiveness distinction is important new learn-ing. I had always taken pride in having chosen field experimentation as the researchmethod that balances internal and external validity, optimizing both. The realizationthat external validity does not guarantee applicability is disconcerting. A string ofnegative results as obtained in Phase 2 could be detrimental to one’s self-confidence.How can I preserve my experimenter self-efficacy, built on the successes of Phase1, in the wake of the successive failures of Phase 2? The answer is Phase 3, whichwill involve replications heeding the above recommendations. Phase 3 will consistof more vigorous attempts to find ways to make PLS training work. This will includesearch for moderators that define the limiting conditions for wittingly producingPygmalion effects in ongoing managerial situations. The findings should help unravelthe efficacy-effectiveness mystery. This is the research agenda for the next genera-tion of my students.

Acknowledgments: This article is based on research conducted for masters’ thesesin organizational behavior under the first author’s supervision by the remainingauthors at Tel Aviv University. The first person singular used in the text refers tothe first author. I owe special thanks to Tamar Cohen, who cotrained two of theexperimental workshops with me. This research was partially supported by grantsfrom the Israel Institute of Business Research and from the Lily and AlejandroSaltiel Chair in International Management. The results of the research reportedhere were presented at the Thirteenth Annual Meeting of the Society for Industrialand Organizational Psychology in Dallas, 1998.

NOTE

1. Oral presentation at symposium entitled “Reflections and Directions for LeadershipResearch” at the Fourteenth Annual Meeting of the Society for Industrial and Organi-zational Psychology, Atlanta, May, 1999.

206 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

REFERENCES

Avolio, B. J., & Bass, B. M. (1988). Transformational leadership, charisma and beyond. InJ. G. Hunt, B. R. Baliga, H. P. Dachler, & C. A. Schriesheim (Eds.), Emerging leadershipvistas (pp. 29–50). Lexington, MA: Heath.

Avolio, B. J., & Bass, B. M. (1995). Individual consideration viewed at multiple levels ofanalysis: A multi-level framework for examining the diffusion of transformationalleadership. Leadership Quarterly, 6, 199–218.

Avolio, B. J., & Gibbons, T. C. (1988). Developing transformational leaders: A life spanapproach. In J. A. Conger & R. N. Kanungo (Eds.), Charismatic leadership: The elusivefactor in organizational effectiveness (pp. 276–308). San Francisco: Jossey-Bass.

Babad, E. Y. (1993). Pygmalion—25 years after interpersonal expectations in the classroom.In P. D. Blanck (Ed.), Interpersonal expectations: Theory, research, and applications(pp. 125–153). Cambridge: Cambridge University Press.

Babad, E. Y., Inbar, J., & Rosenthal, R. (1982). Pygmalion, Galatea, and the Golem: Investiga-tions of biased and unbiased teachers. Journal of Educational Psychology, 74, 459–474.

Bargh, J. A., & Chartrand, T. L. (1999). The unbearable automaticity of being. AmericanPsychologist, 54, 462–479.

Barling, J., Weber, T., & Kelloway, E. K. (1996). Effects of transformational leadershiptraining on attitudinal and financial outcomes: A field experiment. Journal of AppliedPsychology, 81, 827–832.

Bass, B. M. (1985). Leadership and performance beyond expectations. New York: Free Press.Bass, B. M. (1998). Transformational leadership: Industrial, military, and educational impact.

Mahwah, NJ: Lawrence Erlbaum.Bennis, W. G., & Nanus, B. (1985). Leaders: The strategies for taking charge. New York:

Harper and Row.Burke, M. J., & Day, R. R. (1986). A cumulative study of the effectiveness of managerial

training. Journal of Applied Psychology, 71, 232–245.Campbell, D., & Stanley, J. (1966). Experimental and quasi-experimental designs for research.

Chicago: Rand McNally.Caplan, R. D., Vinokur, A. D., Price, R. H., & van Ryn, M. (1989). Job seeking, reemployment,

and mental health: A randomized field experiment in coping with job loss. Journal ofApplied Psychology, 74, 759–769.

Chaiken, S., & Trope, Y. (Eds.). (1999). Dual-process theories in social psychology. NewYork: Guilford Press.

Chapman, G. B., & McCauley, C. (1993). Early career achievements of National ScienceFoundation (NSF) graduate applicants: Looking for Pygmalion and Galatea effects inNSF winners. Journal of Applied Psychology, 78, 815–820.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale,NJ: Lawrence Erlbaum.

Conger, J. A. (1989). The charismatic leader: Behind the mystique of exceptional leadership.San Francisco: Jossey-Bass.

Conger, J. A. (1992). Learning to lead: The art of transforming managers into leaders. SanFrancisco: Jossey-Bass.

Crawford, K. S., Thomas, E. D., & Fink, J. J. (1980). Pygmalion at sea: Improving the workeffectiveness of low performers. Journal of Applied Behavioral Science, 16, 482–505.

Davidson, O. B., & Eden, D. (in press). Remedial self-fulfilling prophecy: Two field experi-ments to prevent Golem effects among disadvantaged women. Journal of AppliedPsychology.

Dawson, J. E., Messe, L. A., & Phillips, J. L. (1972). Effect of instructor-leader behavior onstudent performance. Journal of Applied Psychology, 56, 369–376.

Implanting Pygmalion 207

Dusek, J. B., Hall, V. C., & Meyer, W. J. (Eds.). (1985). Teacher expectations. Hillsdale, NJ:Lawrence Erlbaum.

Dvir, T., Eden, D., & Banjo, M. L. (1995). Self-fulfilling prophecy and gender: Can womenbe Pygmalion and Galatea? Journal of Applied Psychology, 80, 253–270.

Dvir, T., Eden, D., Avolio, B., & Shamir, B. (1999). The impact of transformational leadershiptraining on follower development and performance: A field experiment. Paper presentedat the Fourteenth Annual Meeting of the Society for Industrial and OrganizationalPsychology, Atlanta.

Eden, D. (1984). Self-fulfilling prophecy as a management tool: Harnessing Pygmalion.Academy of Management Review, 9, 64–73.

Eden, D. (1985). Team development: A true field experiment employing three levels ofrigor. Journal of Applied Psychology, 70, 94–100.

Eden, D. (1986). OD and self-fulfilling prophecy: Boosting productivity by raising expecta-tions. Journal of Applied Behavioral Science, 22, 1–13.

Eden, D. (1988a). Creating expectation effects in OD: Applying self-fulfilling prophecy.Research in Organizational Change and Development, 2, 235–267.

Eden, D. (1988b). Pygmalion, goal setting, and expectancy: Compatible ways to raise produc-tivity. Academy of Management Review, 13, 639–652.

Eden, D. (1990a). Consultant as Messiah: Applying expectation effects in managerial consul-tation. Consultation, 9, 37–50.

Eden, D. (1990b). Industrialization as a self-fulfilling prophecy: The role of expectations indevelopment. International Journal of Psychology, 25, 871–886.

Eden, D. (1990c). Pygmalion in management: Productivity as a self-fulfilling prophecy. Lexing-ton, MA: Lexington Books.

Eden, D. (1990d). Pygmalion without interpersonal contrast effects: Whole groups gain fromraising manager expectations. Journal of Applied Psychology, 75, 394–398.

Eden, D. (1993a). Interpersonal expectancy effects in organizations. In P. D. Blanck (Ed.),Interpersonal expectations: Theory, research, and applications (pp. 154–178). Cam-bridge: Cambridge University Press.

Eden, D. (1993b). Leadership and expectations: Pygmalion effects and other self-fulfillingprophecies in organizations. Leadership Quarterly, 3, 271–305.

Eden, D. (1996). From self-efficacy to means efficacy: Internal and external sources of generaland specific efficacy. Paper presented at the 56th Annual Meeting of the Academy ofManagement (Organizational Behavior Division), Cincinnati, Ohio. Published in theSupplemental Electronic Proceedings (SEP) Disk of best papers.

Eden, D. (1999). Means efficacy: External sources of general and specific subjective efficacy.In M. Erez, U. Kleinbeck, & H. Thierry (Eds.), Work motivation in the context of aglobalizing economy. Hillsdale, NJ: Lawrence Erlbaum.

Eden, D., & Aviram, A. (1993). Self-efficacy training to speed reemployment: Helping peopleto help themselves. Journal of Applied Psychology, 78, 352–360.

Eden, D., & Kinnar, J. (1991). Modeling Galatea: Boosting self-efficacy to increase volunteer-ing. Journal of Applied Psychology, 76, 770–780.

Eden, D., & Moriah, L. (1996). Impact of internal auditing on branch bank performance: Afield experiment. Organizational Behavior and Human Decision Processes, 68, 262–271.

Eden, D., & Ravid, G. (1982). Pygmalion versus self-expectancy: Effects of instructor- andself-expectancy on trainee performance. Organizational Behavior and Human Perfor-mance, 30, 351–364.

Eden, D., & Shani, A. B. (1982). Pygmalion goes to boot camp: Expectancy, leadership, andtrainee performance. Journal of Applied Psychology, 67, 194–199.

208 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Eden, D., & Zuk, Y. (1995). Seasickness as a self-fulfilling prophecy: Raising self-efficacyto boost performance at sea. Journal of Applied Psychology, 80, 628–635.

Fiedler, F. E. (1967). A theory of leadership effectiveness. New York: McGraw-Hill.Fiedler, F. E. (1996). Research on leadership selection and training: One view of the future.

Administrative Science Quarterly, 41, 241–250.Geller, D. (1993). Pygmalion goes to work: Developing Pygmalion leadership style as a tool

for improving productivity. Unpublished master’s thesis, Faculty of Management, TelAviv University, Tel Aviv.

Gewirtz, A. (1991). Pygmalion goes to summer camp: Enhancing goal attainment of campersthrough leader expectation training. Unpublished master’s thesis, Department of Psy-chology, Tel Aviv University, Tel Aviv.

Gordon-Terner, R. (1994). Effect of managers’ expectations at two organizational levels ontheir leadership, self-efficacy, and organizational climate. Unpublished master’s thesis,Faculty of Management, Tel Aviv University, Tel Aviv.

Harris, E. F., & Fleishman, E. A. (1955). Human relations training and the stability ofleadership patterns. Journal of Applied Psychology, 39, 20–25.

Harris, M. J., & Rosenthal, R. (1985). Mediation of interpersonal expectancy effects: 31meta-analyses. Psychological Bulletin, 97, 363–386.

Hersey, P., & Blanchard, K. (1982). Management of organizational behavior: Utilizing humanresources (4th ed.). Englewood Cliffs, NJ: Prentice-Hall.

Holon, S. D. (1996). The efficacy and effectiveness of psychotherapy relative to medications.American Psychologist, 51, 1025–1030.

House, R. J. (1977). A 1976 theory of charismatic leadership. In J. G. Hunt & L. L. Larson(Eds.), Leadership: The cutting edge (pp. 189–207). Carbondale, IL: Southern IllinoisUniversity Press.

House, R. J., & Aditya, R. M. (1997). The social scientific study of leadership: Quo vadis?Journal of Management, 23, 409–473.

House, R. J., & Howell, J. M. (1992). Personality and charismatic leadership. LeadershipQuarterly, 3, 81–108.

House, R. J., & Shamir, B. (1993). Toward the integration of transformational, charismatic,and visionary theories. In M. M. Chemers & R. Ayman (Eds.), Leadership theory andresearch: Perspectives and directions. San Diego, CA: Academic Press.

Inbar, I. (1994). Pygmalion skill training among managers as a means for improving workers’self-efficacy, motivation, collective efficacy and climate. Unpublished master’s thesis,Faculty of Management, Tel Aviv University, Tel Aviv.

Jussim, L. (1991). Social perception and social reality: A reflection-construction model.Psychological Review, 98, 54–73.

Jussim, L., & Eccles, J. (1995). Naturally occurring interpersonal expectancies. Review ofPersonality and Social Psychology, 15, 74–108.

Kierein, N., & Gold, M. A. (in press). Pygmalion in organizations: A meta-analysis. Journalof Organizational Behavior.

King, A. S. (1971). Self-fulfilling prophecies in training the hard-core: Supervisors’ expecta-tions and the underprivileged workers’ performance. Social Science Quarterly, 52,369–378.

King, A. S. (1974). Expectation effects in organization change. Administrative Science Quar-terly, 19, 221–230.

Klein, K. J., Dansereau, F., & Hall, R. J. (1994). Levels issues in theory development, datacollection, and analysis. Academy of Management Review, 19, 195–229.

Learman, L. A., Avorn, J., Everitt, D. E., & Rosenthal, R. (1990). Pygmalion in the nursing

Implanting Pygmalion 209

home: The effects of caregiver expectations on patient outcomes. Journal of the Ameri-can Geriatrics Society, 38, 797–803.

Liberman, M. (1991). The effects of expectancy training among commanders in the IsraelDefense Forces on their subordinates’ performance. Unpublished master’s thesis, Fac-ulty of Management, Tel Aviv University, Tel Aviv.

Madon, S., Jussim, L., & Eccles, J. (1997). In search of the powerful self-fulfilling prophecy.Journal of Personality and Social Psychology, 72, 791–809.

McNatt, D. B. (2000). Ancient Pygmalion joins contemporary management: A meta-analysisof the result. Journal of Applied Psychology, 85, 314–322.

Merton, R. K. (1948). The self-fulfilling prophecy. Antioch Review, 8, 193–210.Miller, S. (1986). The impact of Jewish education on the religious behavior and attitudes of

British secondary school pupils. Studies in Jewish Education, 3, 56–67.Mowday, R. T., Steers, R. M., & Porter, L. W. (1979). The measurement of organizational

commitment. Journal of Vocational Behavior, 14, 224–247.Natanovich, G., & Eden, D. (1999). Creating Pygmalion effects among supervisors of outreach

tutors: Replication clarifying gender and ethnic differences (Working Paper No. 13/99).Tel Aviv: Faculty of Management, Tel Aviv University.

Oz, S., & Eden, D. (1994). Restraining the Golem: Boosting performance by changing theinterpretation of low scores. Journal of Applied Psychology, 79, 744–754.

Pass, Y. (1994). Effect of Pygmalion management style of school headmasters. Unpublishedmaster’s thesis, Faculty of Management, Tel Aviv University, Tel Aviv.

Raudenbush, S. W. (1984). Magnitude of teacher expectancy effects on pupil IQ as a functionof the credibility of expectancy induction: A synthesis of findings from 18 experiments.Journal of Educational Psychology, 76, 85–97.

Rosenthal, R. (1979). The “file drawer problem” and tolerance for null results. PsychologicalBulletin, 86, 638–641.

Rosenthal, R. (1985). From unconscious experimenter bias to teacher expectancy effects. InJ. B. Dusek, V. C. Hall, & W. J. Meyer (Eds.), Teacher expectations (pp. 37–65).Hillsdale, NJ: Lawrence Erlbaum.

Rosenthal, R. (1991a). Meta-analytic procedures for social science. Newbury Park, CA: Sage.Rosenthal, R. (1991b). Teacher expectancy effects: A brief update 25 years after the Pygmal-

ion experiment. Journal of Research in Education, 1, 3–12.Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bulletin, 118, 183–192.Rosenthal, R., & Jacobson, L. (1968). Pygmalion in the classroom: Teacher expectation and

pupils’ intellectual development. New York: Holt, Rinehart and Winston.Rosenthal, R., & Rubin, D. B. (1978). Interpersonal expectancy effects: The first 345 studies.

Behavioral and Brain Studies, 3, 377–386.Rosenthal, R., & Rubin, D. B. (1982). A simple general purpose display of magnitude of

experimental effect. Journal of Educational Psychology, 74, 166–169.Sashkin, M. (1988). The visionary leader. In J. A. Conger & R. N. Kanungo (Eds.), Charismatic

leadership: The elusive factor in organizational effectiveness (pp. 122–160). San Fran-cisco: Jossey-Bass.

Salomon-Segev, I. (1994). Pygmalion in bank branches: Boosting managerial effectivenessusing a Pygmalion-based leadership training workshop. Unpublished master’s thesis,Faculty of Management, Tel Aviv University, Tel Aviv.

Seligman, M. E. P. (1995). The effectiveness of psychotherapy: The Consumer Reports study.American Psychologist, 50, 965–974.

Seligman, M. E. P. (1996). Science as an ally of practice. American Psychologist, 51, 1072–1079.Seligman, M. E. P., & Levant, R. F. (1998). Managed care policies rely on inadequate science.

Professional Psychology: Research & Practice, 29, 211–212.

210 LEADERSHIP QUARTERLY Vol. 11 No. 2 2000

Shalit, M. (1996). The effect of organizational interventions on business performance: Twofield experiments on Pygmalion workshops. Unpublished master’s thesis, Faculty ofManagement, Tel Aviv University, Tel Aviv.

Sherer, M., Maddux, J. E., Mercadante, B., Prentice-Dunn, S., Jacobs, B., & Rogers, R. W.(1982). The Self-efficacy Scale: Construction and validation. Psychological Reports,51, 663–671.

Smith, A. E., Jussim, L., Eccles, J., VanNoy, M., Madon, S., & Palumbo, P. (1998). Self-fulfilling prophecies, perceptual biases, and accuracy at the individual and group levels.Journal of Experimental Social Psychology, 34, 530–561.

Smoll, F. L., Smith, R. E., Barnett, N. P., & Everett, J. J. (1993). Enhancement of children’sself-esteem through social support training for youth sport coaches. Journal of AppliedPsychology, 78, 602–610.

Terpstra, D. (1981). Relationship between methodological rigor and reported outcomes inorganization development evaluation research. Journal of Applied Psychology, 66,541–543.

Vecchio, R. P. (Ed.). (1997). Leadership: Understanding the dynamics of power and influencein organizations. Notre Dame: University of Notre Dame Press.

Vroom, V. H., & Jago, A. G. (1988). The new leadership: Managing participation in organiza-tions. Englewood Cliffs, NJ: Prentice-Hall.

Weinstein, R. S., Soule, C. R., Collins, F., Cone, J., Mehlhorn, M., & Simontacchi, K. (1991).Expectations and high school change: Teacher–researcher collaboration to preventschool failure. American Journal of Community Psychology, 19, 333–363.

Yukl, G. (1981). Leadership in organizations (1st ed.). Englewood Cliffs, NJ: Prentice-Hall.Yukl, G. (1998). Leadership in organizations (4th ed.). Englewood Cliffs, NJ: Prentice-Hall.