Pygmalion effect

type: effect · origin: study · evidence: supported · themes: how we judge others, judgements and probabilities

Claim assessed: Teachers' expectations change pupils' school results.

Published 8 October 2026 · how we work

In short

The Pygmalion effect is the idea that someone’s expectations about you can come true simply because they hold them. In the famous version, teachers who were told that some pupils would “bloom” intellectually treated them differently, and those pupils actually became more intelligent. The effect exists, but it is much smaller than in this story. On average, about 5–10% of pupils are affected, the effect shrinks once the teacher knows the pupils, and teachers’ expectations predict results mostly because they are accurate. The claim that expectations raise children’s intelligence has not held up.

What it says

The original experiment. Rosenthal and Jacobson (1968) gave every child in an American primary school a nonverbal intelligence test, presented to the teachers as a new Harvard test that shows which children will make an intellectual leap in the coming year. They then gave the teachers the names of these “bloomers”, about 20% of the school, actually picked at random. After a year, both groups had gained IQ points: the “bloomers” 12 on average, the others 8. The difference, about 4 points, was significant but small (a correlation of 0.15). The effect appeared almost only in the first two grades, with differences of 15 and 10 points; in 8 of the 11 comparisons by grade and year it did not appear. We haven’t read the book; we describe it from Jussim and Harber (2005) and Spitz (1999).

The data problems. According to Jussim and Harber, Snow (1995) reanalysed the data and found impossible scores among the youngest children: starting IQs near zero and final ones above 200. The test was normed only for scores between 60 and 160. If you drop the scores outside that range, the effect disappears. It also disappears if you drop five children with average gains of over 90 points, for example from 17 to 110. Spitz (1999) tells the history of the controversy: the three replications that told teachers, as in the original, that some children would “bloom” intellectually did not reproduce the result.

The meta-analyses. Rosenthal and Rubin (1978) collected 345 experiments on how one person’s expectations affect another’s behaviour and concluded that the effect is real. For IQ, Raudenbush (1984) combined 18 experiments in which teachers were given false expectations. The average effect was very small (d = 0.11), and 8 of the 19 effects came out negative. The size of the effect, however, depended on how well the teachers knew their pupils when they got the false information. We describe both from Jussim and Harber and from Spitz; the articles are available only behind a paywall.

Expectations in real life. Most studies don’t lie to teachers; they measure their expectations at the start of the year and follow the pupils, taking their earlier grades and results into account. Jussim and Harber (2005), in a review of 35 years of research, conclude that:

• the effects are real but usually small (correlations of 0.1–0.2), which means they change the results of about 5–10% of pupils;
• teachers’ expectations predict results mostly because they are accurate; by their estimate, about three quarters of the link between expectations and test results is accuracy, not prophecy;
• the effects don’t add up from year to year but fade, though they don’t vanish completely: in the data of Smith, Jussim and Eccles (1999), the effect of sixth-grade expectations on test scores fell from 0.08 in seventh grade to 0.02 in twelfth grade;
• one study (Jussim et al., 1996) found large effects, of 0.4–0.6, among African American pupils and of about 0.6 among low-achieving pupils from poor families; each of these group differences, the authors note, was tested in a single study.

Recent large studies.

• Gentrup et al. (2020) followed 1,026 first-graders in 64 classes in Germany. The expectations teachers held at the start, beyond what the children’s results, abilities and motivation justified, predicted results at the end of the year: in reading, β = 0.21, with 4% more of the variation explained; in maths, β = 0.09, with under 1% more. Teachers’ feedback, filmed in class, did not explain the link. The study is correlational.
• Papageorge, Gershenson and Kang (2020) used data on 6,060 American tenth-graders for whom two different teachers said how far they thought the pupil would go in school. They compared what happens when the two disagree. An expectation 10–20 percentage points higher that the pupil would finish college raised the chance that they actually did by 1.4–2.8 percentage points, from a base of 45%. The authors call the effect “modest, but nontrivial”.

Interventions. De Boer, Timmermans and van der Werf (2018) collected 19 programmes that tried to raise teachers’ expectations. According to the abstract, the programmes raised expectations (g = 0.38) and then pupils’ results (g = 0.30). Rubie-Davies and Hattie (2024) describe large differences (d ≈ 0.9) between the classes of high- and low-expectation teachers, but from often small studies in which teachers were split into extreme groups.

Example

A teacher takes on a new class at the start of the year and learns from the records that one pupil “has problems”. Without noticing, she calls on him less often, gives him easier tasks and praises a good answer less. By the end of the year, the pupil has fallen further behind than he otherwise would have. The data say this can happen, but also that, for most pupils, the teacher’s view would have been largely right from the start. The example is an illustration, not a study cited here.

How to apply it

The steps below are a practical way to apply it that we propose, based on the studies cited.

1. Watch the first weeks. The effect of false expectations was largest when teachers didn’t yet know their pupils. That is when it matters most not to draw conclusions from records or from what others say.
2. Check your impression against data. Expectations that match the pupil’s actual results don’t produce the effect. Those that drift away from them, up or down, can.
3. Give everyone the same chances. Studies of teachers who treat strong and weak pupils very differently show larger effects. Questions, hard tasks and feedback for everyone leave less room for an expectation to slip in.
4. Don’t rely on it as a method. Telling yourself a child is a genius doesn’t make them one. The effect is small and appears only in some pupils.

Limits and nuances

• Intelligence doesn’t rise from expectations. The original claim, that teachers’ expectations raise IQ, remains somewhere between nil (according to the critics) and very small (according to the supporters). Jussim and Harber note that only this much is certain: a large effect on intelligence has been disconfirmed.
• Correlational, not experimental. The recent large studies don’t manipulate expectations; they measure them. Even controlling for earlier results, part of the link may come from what the teacher knew about the pupil that wasn’t measured. Papageorge et al. try to solve this by comparing two teachers, but they too stop at “arguably causal”.
• Positive or negative? It is unclear whether low expectations harm more than high ones help. Jussim and Harber find only three studies on this question, with contradictory results.
• At work. McNatt (2000), across 17 studies, found a large effect (d = 1.13), especially in the military. Eden et al. (2000), however, trained managers in a “Pygmalion style” in seven field experiments and found a small effect (r = 0.13) and little evidence that the training changed anything. We know both only from their abstracts.
• We didn’t read a meta-analysis in full. Our rule is to read at least one meta-analysis or large replication. For the Pygmalion effect, the meta-analyses (Rosenthal and Rubin 1978, Raudenbush 1984, McNatt 2000, de Boer et al. 2018) are not open access or we couldn’t reach their text. The entry rests on Jussim and Harber’s review and on two large studies, Gentrup et al. and Papageorge et al. We found no preregistered replication and no meta-analysis corrected for unpublished studies. It is an exception to the rule, which we flag here.
• What we read. We read in full Jussim and Harber (2005) and Spitz (1999), from copies on gwern.net, Gentrup et al. (2020), from the University of Bamberg repository, Papageorge et al., in the 2018 NBER working paper version, and Rubie-Davies and Hattie (2024), from PubMed Central. Rosenthal and Jacobson’s book, Rosenthal and Rubin (1978) and Raudenbush (1984) we describe from Jussim and Harber and from Spitz. McNatt (2000), Eden et al. (2000) and de Boer et al. (2018) we read only as abstracts; the last one’s page blocked our access, although the article is open access. De Boer et al.’s meta-analysis could show how solid the interventions are, but with 19 programmes it is unlikely to change the evidence level.

Sources

• Robert Rosenthal, Lenore Jacobson (1968). Pygmalion in the classroom: Teacher expectation and pupils' intellectual development
• Robert Rosenthal, Donald B. Rubin (1978). Interpersonal expectancy effects: The first 345 studies
• Stephen W. Raudenbush (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
• Herman H. Spitz (1999). Beleaguered Pygmalion: A history of the controversy over claims that teacher expectancy raises intelligence
• Brian McNatt (2000). Ancient Pygmalion joins contemporary management: A meta-analysis of the result
• Dov Eden et al. (2000). Implanting Pygmalion leadership style through workshop training: Seven field experiments
• Lee Jussim, Kent D. Harber (2005). Teacher expectations and self-fulfilling prophecies: Knowns and unknowns, resolved and unresolved controversies
• Hester de Boer, Anneke C. Timmermans, Margaretha P. C. van der Werf (2018). The effects of teacher expectation interventions on teachers' expectations and student achievement: Narrative review and meta-analysis
• Sarah Gentrup, Georg Lorenz, Cornelia Kristen, Irena Kogan (2020). Self-fulfilling prophecies in the classroom: Teacher expectations, teacher feedback and student achievement
• Nicholas W. Papageorge, Seth Gershenson, Kyung Min Kang (2020). Teacher expectations matter
• Christine M. Rubie-Davies, John A. C. Hattie (2024). The powerful impact of teacher expectations: A narrative review

See also: Confirmation bias, Halo effect, Hawthorne effect, Stereotype threat