the Pygmalion effect: what other people's expectations really do to performance

Before the classroom: the rats

Robert Rosenthal, who spent much of his career at Harvard and died in 2024 at the age of 90, began with a question about researchers themselves. Could an experimenter's expectations change the results of an experiment?

In a study published in 1963 with Kermit Fode, twelve students in an experimental psychology course were each given five rats to train on a simple T-shaped maze. Half the students were told their rats came from a line bred to be "maze-bright", and half that theirs were "maze-dull". In fact, the rats had been assigned to the labels at random. Over five days of trials, the rats run by students who expected bright animals made more correct choices on average than those run by students who expected dull ones.

The sample was small and the differences were modest, but the point landed: people can pass on their expectations without meaning to. Harvard's memorial tribute notes that this line of work helped push psychologists toward double-blind experiments, where the person running the study does not know which group is which.

The study: Oak School

Rosenthal then asked whether the same thing happened in classrooms. Working with Lenore Jacobson, he ran the study that became the 1968 book Pygmalion in the Classroom.

According to the account in Robert Thorndike's review of the book, children from kindergarten through the sixth grade at an elementary school took a group intelligence test in May 1964. Teachers were told the test could predict which children were likely to show an academic spurt. That September, each teacher received a list of names of these supposed "bloomers". The names had been chosen using random numbers. In a later summary, Stephen Raudenbush describes the bloomers as a randomly selected 20 percent of the pupils. The children were retested in January 1965, May 1965 and May 1966.

What it found

Rosenthal and Jacobson reported that, after about eight months, the "bloomers" had gained significantly more than their classmates: close to 4 points more in total IQ and 7 points more in reasoning IQ, according to Raudenbush's summary. Teachers also rated them as more curious and better adjusted.

The message that reached the public was simple and hopeful: believe in a child, and the child will rise.

Did it hold up?

The original study drew sharp criticism. Robert Thorndike reviewed the book in the American Educational Research Journal in 1968. He called it "so defective technically" that he regretted it had ever left the original investigators. Looking at the authors' own tables, he pointed out that the advantage appeared almost entirely in the first and second grades, among 19 children. In the other four grades the differences were near zero. He also showed that some of the test scores could not be right: one class of 19 young pupils had a reported average reasoning "IQ" of about 31 before the study began, which would mean the test was not working at all for the youngest children.

Richard Snow's 1969 review in Contemporary Psychology added more problems. The test had no proper norms for the youngest children, some individual scores swung implausibly between testings, and the reporting made it hard to check the analysis.

Later experiments gave a more modest picture. In 1984, Raudenbush pooled 18 experiments that had tried to raise pupils' IQ by giving teachers high expectations. The average effect was small, about a tenth of a standard deviation, and only five of the 18 studies reached statistical significance. But one pattern stood out. When teachers had no contact with the pupils before being given the expectation, the average effect was about 0.32. After one week of contact it was about 0.26. When teachers had known the pupils for more than two weeks, the effect disappeared. Effects were also larger in the first two grades and, unexpectedly, again in grade 7. In other words, a planted expectation mattered mainly when teachers had little else to go on.

The overall verdict: real, but small. In a 2005 review of 35 years of research in Personality and Social Psychology Review, Lee Jussim and Kent Harber concluded that self-fulfilling prophecies in classrooms do happen, but are typically small, tend not to build up over time, and may fade. They may be stronger for students from stigmatised groups. They also noted that teachers' expectations often predict how students do mainly because the expectations are accurate, not because they cause the result. Jussim's own view is that the effect on IQ in particular remains, at best, weakly established.

How it gets misused

• "Expect more and they will deliver." The research shows expectations can tilt outcomes slightly, mostly early on. It does not show that belief alone transforms anyone.
• Quoting the IQ result as fact. This is the most disputed part of the whole story.
• The reverse blame. If a learner struggles, it does not follow that their teacher, manager or parent did not believe in them enough. Most of the time, expectations follow evidence.

What it means for you

The careful lesson is not about belief. It is about how expectations get communicated, especially at the start, when someone new joins your team, your class or your family's routine and you have little else to go on.

Early judgements come out in small ways: who you give the stretching task to, how long you wait for an answer, how much detail you put in your feedback, whose ideas you build on. A manager in Pune who has decided a new hire is "not client-facing" may stop inviting her onto client calls, and then she never gets the practice that would prove him wrong. That is not magic. It is opportunity, and it is something you can check.

Say high expectations in a way that is specific, and pair them with support:

"This report is harder than the last one. I am giving it to you because your analysis on the vendor review was sharp. Come to me on Thursday with a first draft and we will work on it together."

At home, the same principle applies to a teenager preparing for exams or a younger child learning to read:

"I have seen how you work when something matters to you. What would help you most this week?"

And watch for the quieter version of a low expectation:

"Let me just do it, it will be quicker."

Sources

1. Harvard Gazette, Memorial Minute for Robert Rosenthal, 90, October 2024. news.harvard.edu/gazette/story/2024/10/memorial-minute-for-robert-rosenthal-90
2. Robert Rosenthal and Kermit L. Fode, The Effect of Experimenter Bias on the Performance of the Albino Rat, Behavioral Science 8(3), 183-189, 1963 (scanned copy hosted by gwern.net). www.gwern.net/doc/iq/animal/1963-rosenthal.pdf
3. Robert L. Thorndike, review of Pygmalion in the Classroom by Rosenthal and Jacobson, American Educational Research Journal 5(4), 708-711, November 1968 (scanned copy hosted by gwern.net). www.gwern.net/doc/iq/1968-thorndike.pdf
4. Richard E. Snow, Unfinished Pygmalion, Contemporary Psychology 14(4), 197-199, 1969 (copy hosted by gwern.net). gwern.net/doc/iq/1969-snow.html
5. Stephen W. Raudenbush, 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, 1984 (scanned copy hosted by gwern.net). gwern.net/doc/statistics/bias/1984-raudenbush.pdf
6. Lee Jussim and Kent D. Harber, Teacher Expectations and Self-Fulfilling Prophecies: Knowns and Unknowns, Resolved and Unresolved Controversies, Personality and Social Psychology Review 9(2), 131-155, 2005. doi.org/10.1207/s15327957pspr0902_3
7. Lee Jussim, Chapter 3. The Once Raging and Still Smoldering Pygmalion Controversy, excerpt on the author's Rutgers site. sites.rutgers.edu/lee-jussim/?p=599

Published by The Full Stop, the journal of Brightside English, in the The Experiments series. Every factual claim links to a source below; where accounts or findings are disputed, we say so. Sources checked on 26 Oct 2026. Reviewed by a Brightside senior coach. Corrections: see our editorial standard.