Does the Pygmalion Effect Really Work
Adam Alter's Drunk Tank Pink reports the Oak School result at full strength. Later experiments find a smaller effect that fades once teachers know their pupils, and larger ones in army platoons and workplaces.
Answer to the question
Does the Pygmalion effect really work?
Yes, in a small and conditional form. Across eighteen classroom experiments the average Pygmalion effect on pupil IQ was small on average, and gone where teachers already knew their pupils. Larger effects appear in army and workplace studies, in unusual settings.
Key takeaways
- Rosenthal and Jacobson (1966, 1968) told teachers that randomly chosen pupils were “academic bloomers”. Jussim and Harber (2005) put the whole-school difference at about four IQ points, significant mainly in the first two grades.
- Thorndike (1968) found implausible pretest scores, including a control class of nineteen with an average reasoning “IQ” of about 31, and judged that the data could not carry the conclusion.
- Raudenbush (1984) pooled 18 experiments. The average effect was small, larger where teachers had known pupils two weeks or less, and gone beyond that.
- Eden (1990) randomly assigned 29 Israeli army platoons. McNatt (2000) pooled 17 adult studies with 2,874 people and found an average far larger than in classrooms, strongest in military settings.
- Rubie-Davies and colleagues (2015) randomly assigned 84 teachers to training in high-expectation practices. Their pupils improved more in mathematics over one year.
Robert Thorndike read the book that launched the Pygmalion effect in 1968 and could not make its numbers work. One class of nineteen pupils had a control-group reasoning "IQ" of about thirty-one. The first-grade pretest, by his arithmetic, implied an average raw score of about two on a test where random marking would earn five or six.
Most readers have never met that review. They meet the result in retellings, and the cleanest is in Adam Alter's Drunk Tank Pink. Robert Rosenthal and Lenore Jacobson told the teachers at a San Francisco elementary school, which Alter calls Oak School, that certain pupils were "academic bloomers" about to show a spurt of intellectual growth. The pupils had been chosen at random. After a summer break and a retest, Alter reports, the first and second graders among them had outscored their classmates by ten to fifteen IQ points, and four in five of the bloomers had gained at least ten points, against half of the others. The book gives the result at full strength.
What the Oak School Numbers Leave Out of the Headline
The figure Alter prints is the result for the two youngest grades. Lee Jussim and Kent Harber, reviewing thirty-five years of research in 2005, put the average difference across the whole school at about four IQ points, and found it significant in only some grades, mainly the first and second. Both groups gained over the year, so the Pygmalion effect is the difference between two gains. The Pygmalion effect in the original study is a modest extra gain in a school where nearly every child gained, concentrated among the youngest pupils.
Thorndike's review, in the American Educational Research Journal, did not dispute that an expectation can help bring about what it predicts. His objection was that the data could not carry the conclusion, so that if the authors were right, it was by good fortune. The report gave no raw scores and no pupil ages, which forced him to estimate the underlying figures from the test's published norms.
Alter explains the result through the teachers, who emphasised the bloomers' strengths and read their mistakes as flukes. That is a plausible mechanism, and the study's outcome was test scores, so it recorded no classroom behaviour that would confirm it.
Eighteen Experiments, a Two-Week Rule and a Rival Explanation
Other researchers tried the same manipulation, and Stephen Raudenbush pooled eighteen of those experiments in 1984. The average Pygmalion effect on pupil IQ was small. The pattern behind that average mattered more than the average. Where teachers had known their pupils for two weeks or less when the researchers handed them the label, the effect was larger, though still modest. Where they had known them longer, it disappeared, and the average for those experiments was slightly below zero. Raudenbush's abstract says the data strongly supported the idea that the better teachers know their pupils at the time of the label, the smaller the effect.
Effects were also larger in grades one and two than in grades three to six, and they reappeared in grade seven. Raudenbush's abstract reports no support for two other explanations, that the type of IQ test mattered or that the testers' awareness of the labels did.
The result suggests that the Pygmalion effect needs a teacher who has no view of her own yet. Once a teacher has watched a child for a few weeks, the researcher's list adds little. For the Pygmalion effect in an ordinary classroom, which is mostly populated by teachers who already know their pupils, that finding does most of the damage.
Jussim and Harber raise a second problem, which concerns prediction in ordinary life and not the lab. Teachers' expectations do predict how pupils later perform, which is why the Pygmalion effect is hard to separate from accurate judgement. The question is whether the expectations caused the performance or simply read it correctly.
Their conclusion was that expectations in classrooms do sometimes bring about what they predict, that the Pygmalion effect is typically small, and that it does not build up much across teachers or over time. By their account expectations alter the achievement of roughly five to ten per cent of pupils, so most pupils are unaffected. Effects may be stronger for pupils from stigmatised groups. Oak School tested only positive labels, so it said nothing about whether negative expectations do harm, and whether any of this reaches intelligence itself, or does more harm than good, remains unresolved in their account.
Do Platoons and Workplaces Show Larger Effects?
Outside schools the Pygmalion effect is reported as larger, and the settings are unusual. Dov Eden randomly assigned twenty-nine platoons in the Israel Defense Forces to a Pygmalion condition, in which platoon leaders were told their trainees had unusually high command potential, or to a control condition. The Pygmalion platoons outperformed the controls, and because whole platoons were treated, no trainee was singled out against a comrade. Eden's point was that earlier studies could have been inflated by contrast, in which a leader who is told that some people are special treats the others worse. Anything a leader did differently reached everyone.
Brian McNatt's 2000 meta-analysis of adult settings pooled seventeen studies, with 2,874 participants in all, and reported an average effect far larger than any seen in classrooms. The outcomes ranged from exam scores to performance appraisals to physical output. It was stronger in the military, among men, and where people started with low expectations placed on them, and McNatt noted that ongoing work settings and for-profit firms had been studied far less than training courses.
David Robson, writing in The Expectation Effect, treats the Israeli figures as an outlier and says as much. The Israeli effect is roughly ten times the classroom average, which is one reason to read the workplace and military results as describing particular conditions and not a general rate.
| Study | Setting | What was done | Result as reported |
|---|---|---|---|
| Rosenthal and Jacobson (1966, 1968) | One California elementary school | Teachers told random pupils were bloomers | About four IQ points school-wide, strongest in grades one and two |
| Raudenbush (1984) | Eighteen classroom experiments | Teachers given induced expectations | Small on average, larger with brief contact, gone after two weeks |
| Eden (1990) | Twenty-nine Israeli army platoons, randomised | Leaders told trainees had high potential | Pygmalion platoons outperformed controls |
| McNatt (2000) | Seventeen studies of adults, 2,874 people | Meta-analysis of induced expectations | Far larger than in classrooms, strongest in the military |
| Rubie-Davies and colleagues (2015) | Teachers randomly assigned to training or not | Four workshops on high-expectation practices | Intervention pupils improved more in mathematics over a year |
The Teachers Who Were Trained and Not Told
The strongest classroom evidence comes from a design that gives up the original trick. Christine Rubie-Davies and colleagues randomly assigned eighty-four teachers in New Zealand to an intervention or a control group. Intervention teachers attended four workshops on the practices of high-expectation teachers and planned changes to their own classrooms, and researchers visited three more times to check that the changes were happening. The project's own description names flexible grouping in place of ability grouping, a warm class climate, goal-setting, feedback and nonverbal signals of expectation among the practices taught. Over one year their pupils improved more in mathematics than the pupils of control teachers, and the teachers themselves were highly satisfied with the changes they had made.
Nobody in that trial was handed a false list of gifted children, so it did not repeat the Oak School manipulation. The expectation that moved results there was one the teachers had built themselves, over a year of lessons, and the Pygmalion effect that survives the evidence is of that kind: slow, deliberate and visible in what a teacher does.
Common questions
Does the Pygmalion effect really work?
Yes, in a small and conditional form. Teacher expectations can change pupil outcomes, but the average effect in classroom experiments is small, and it fades once teachers know their pupils.
Was the Oak School study fair?
It was a randomised design with a deceptive label, and it has been criticised since 1968 for unreliable testing and for effects that appear mainly in the first two grades.
Does the Pygmalion effect apply to managers?
Experiments with managers and trainees, mostly in military settings, report larger effects than classrooms do. The effect is stronger when people start with low expectations placed on them.
Sources
Alter, A. (2013). Drunk Tank Pink: And Other Unexpected Forces That Shape How We Think, Feel, and Behave. New York: Penguin Press. Seed book; vault hub Books/Drunk Tank Pink.
Rosenthal, R., & Jacobson, L. (1966). “Teachers' expectancies: Determinants of pupils' IQ gains.” Psychological Reports, 19(1), 115-118. doi:10.2466/pr0.1966.19.1.115 https://doi.org/10.2466/pr0.1966.19.1.115
Thorndike, R. L. (1968). Review of Pygmalion in the Classroom, by R. Rosenthal and L. Jacobson. American Educational Research Journal, 5(4), 708-711. doi:10.3102/00028312005004708 https://doi.org/10.3102/00028312005004708
Raudenbush, S. W. (1984). “Magnitude of teacher expectancy effects on pupil IQ as a function of the credibility of expectancy induction: A synthesis of findings from 18 experiments.” Journal of Educational Psychology, 76(1), 85-97. doi:10.1037/0022-0663.76.1.85.
Jussim, L., & Harber, K. D. (2005). “Teacher expectations and self-fulfilling prophecies: Knowns and unknowns, resolved and unresolved controversies.” Personality and Social Psychology Review, 9(2), 131-155. doi:10.1207/s15327957pspr0902_3 https://doi.org/10.1207/s15327957pspr0902_3
Eden, D. (1990). “Pygmalion without interpersonal contrast effects: Whole groups gain from raising manager expectations.” Journal of Applied Psychology, 75(4), 394-398. doi:10.1037/0021-9010.75.4.394 https://doi.org/10.1037/0021-9010.75.4.394
McNatt, D. B. (2000). “Ancient Pygmalion joins contemporary management: A meta-analysis of the result.” Journal of Applied Psychology, 85(2), 314-322. doi:10.1037/0021-9010.85.2.314 https://doi.org/10.1037/0021-9010.85.2.314
Rubie-Davies, C. M., Peterson, E. R., Sibley, C. G., & Rosenthal, R. (2015). “A teacher expectation intervention: Modelling the practices of high expectation teachers.” Contemporary Educational Psychology, 40, 72-85. doi:10.1016/j.cedpsych.2014.03.003 https://doi.org/10.1016/j.cedpsych.2014.03.003
Robson, D. (2022). The Expectation Effect: How Your Mindset Can Change Your World. Edinburgh: Canongate (the Israeli platoon study treated as an exceptional effect size).