The Oak School Pygmalion Experiment: How Teacher Expectations Change Children

At a California elementary school in 1965, every student was given an IQ test. Teachers were told the test could identify which students were on the verge of a major intellectual growth spurt. The researchers identified about 20 percent of the school's students as "intellectual bloomers." None of those students had actually scored differently on the test — they were chosen at random. But a year later, many of them showed genuine IQ gains that their randomly-not-chosen classmates did not. A teacher's belief, planted deliberately and falsely, had changed how a child's intelligence was measured.

The Research Question and Its Context

The Pygmalion experiment emerged from a specific question in experimental psychology: do expectations function as self-fulfilling prophecies? The idea had roots in sociology, particularly in Robert Merton's 1948 elaboration of the self-fulfilling prophecy concept — the process by which a false belief, by shaping behavior, produces outcomes that make the belief come true after all.

Self-fulfilling prophecy — a situation in which a belief or expectation, held even without basis, produces behaviors that cause the expected outcome to actually occur.

Robert Rosenthal had demonstrated expectancy effects in laboratory animals before he turned to human subjects. In the early 1960s, he found that experimenters who were told their rats were "bright" obtained better maze-learning results than experimenters told their rats were "dull" — even when the rats were randomly assigned to groups and were not actually different in ability. The effect appeared to travel from experimenter belief to rat behavior through the experimenter's handling, timing, and environmental setup.

Source: Rosenthal, R. & Fode, K.L. (1963), "The effect of experimenter bias on the performance of the albino rat," Behavioral Science 8(3), 183–189 — the foundational rat-maze experiment demonstrating experimenter expectancy effects.

The question Rosenthal then asked was whether a similar mechanism could operate in schools: if teachers expected certain students to excel, would those students actually perform better?

Lenore Jacobson, the principal of Oak School, an elementary school in South San Francisco, California, agreed to collaborate. The school served a working-class community with a significant proportion of Latino children.

The Experimental Design

In the spring of 1964, all students at Oak School were administered the Flanagan Tests of General Ability — a standard IQ test. Teachers were told the test was called the "Harvard Test of Inflected Acquisition," a nonexistent instrument specifically invented for the study. They were told the test could identify students who were about to experience unusual intellectual development.

Before the school year began in September 1964, teachers received a list of students who had supposedly scored in the top 20% on this test and who were therefore expected to show significant IQ gains during the coming year. The listed students were selected entirely at random using a random number table. They were not actually different on any measure from their classmates.

Teachers were asked to keep the identifications confidential. The study involved approximately 320 students across 18 classrooms spanning grades one through six.

Source: Rosenthal, R. & Jacobson, L. (1968), Pygmalion in the Classroom: Teacher Expectation and Pupils' Intellectual Development, Holt, Rinehart & Winston — the book reporting the full study, including methodology, results, and interpretations.

Students were retested with the same Flanagan IQ instrument at the end of the first school year (spring 1965) and again at the end of the second school year (spring 1966). Rosenthal and Jacobson then compared IQ gains between the designated "bloomers" and the control students.

The title they chose for the book — Pygmalion in the Classroom — referenced the ancient Greek myth of Pygmalion, a sculptor who fell in love with his own creation, and by extension George Bernard Shaw's 1913 play of the same name, in which a linguistics professor transforms a cockney flower seller into a society lady partly through the force of his expectations and the treatment those expectations produced.

The Results

At the end of the first year, the "bloomers" showed significantly greater gains in IQ scores than the control students, with the effect concentrated in the younger grades. First- and second-grade "bloomers" showed particularly striking gains.

Grade	Bloomer IQ Gain	Control IQ Gain	Difference
1st grade	+12.2 points	+1.8 points	+10.4
2nd grade	+7.0 points	+0.0 points	+7.0
3rd grade	+5.0 points	+5.0 points	0.0
4th grade	+5.6 points	+2.2 points	+3.4
5th grade	+17.4 points	+17.5 points	-0.1
6th grade	+10.7 points	+10.0 points	+0.7

Source: Rosenthal & Jacobson (1968), Pygmalion in the Classroom, Table 7-1 — the full grade-by-grade IQ gain data from the first-year follow-up.

The differences were most pronounced in the younger grades, which Rosenthal and Jacobson interpreted as reflecting greater plasticity in younger children and greater teacher influence in the early grades where children were more dependent on adult guidance.

Rosenthal and Jacobson also collected teacher ratings of the students' academic behavior, adjustment, and intellectual curiosity. Teachers rated "bloomers" as more intellectually curious, more interesting, and better adjusted than control students — a pattern that itself reflects either accurate perception of real change or a halo effect from the expectation.

How the Effect Works: Four Channels

Rosenthal later proposed that teacher expectancy effects operate through four behavioral channels, which he called the four-factor model or, in later presentations, CATE — Climate, Affect, Teaching, and Equity of opportunity.

Climate: Teachers create a warmer social-emotional environment for students they expect to succeed. More eye contact, more smiling, more leaning in, more physical proximity — all of which communicate positive regard.

Feedback: High-expectancy students receive more differentiated feedback on their work. Rather than a general "good job," they receive specific comments that identify what was strong and what could be improved, which supports learning more effectively.

Input: Teachers teach more material, more complex material, and more challenging material to high-expectancy students. They assign harder problems and spend more time explaining difficult concepts.

Output (Equity): Teachers provide more opportunity for high-expectancy students to respond, ask questions, and engage publicly in class. They wait longer after asking a question for a high-expectancy student to answer, rather than moving on.

Source: Rosenthal, R. (1989), "Building a better Pygmalion," Contemporary Psychology — and the four-factor CATE model in Rosenthal's subsequent theoretical writing reviewed in: Harris, M.J. & Rosenthal, R. (1985), "Mediation of interpersonal expectancy effects: 31 meta-analyses," Psychological Bulletin 97(3), 363–386.

Each of these channels individually has measurable effects on learning. Taken together, across an entire school year, they accumulate into real differences in cognitive development.

Think of it like compound interest applied to learning opportunity. A small daily advantage in the quality and quantity of instruction, repeated across 180 school days, produces a meaningfully different trajectory. An IQ test at the end of the year is measuring not just a child's innate capacity but the cumulative effect of how much cognitively challenging input they received and how the adults around them responded to their attempts.

The Controversy

Pygmalion in the Classroom was not greeted with uniform acceptance. The book attracted fierce criticism almost immediately after publication, on multiple grounds.

Statistical critiques were the most specific. Richard Snow, a Stanford educational psychologist, examined the data and pointed out that the IQ gains in the first-grade subsample were extremely large and statistical unusual — the kind of gains that raise questions about data quality. He also noted that the distributions of scores were unusual in ways that suggested possible errors in administration or scoring.

Source: Snow, R.E. (1969), "Unfinished Pygmalion," Contemporary Psychology 14(4), 197–199 — an early and influential methodological critique of the original study.

Replication problems emerged quickly. A number of researchers attempted to replicate the effect and found inconsistent results. Some studies produced expectancy effects similar to Rosenthal and Jacobson's; many did not.

The "bloomers were harmed" controversy was also significant. Some critics noted that in the Rosenthal-Jacobson data, children in the control group who made unexpected IQ gains were rated by teachers as less well-adjusted than control children who did not gain — suggesting that unexpected success from a child not labeled as a bloomer was interpreted negatively by teachers.

Source: Thorndike, R.L. (1968), "Review of Pygmalion in the Classroom," American Educational Research Journal 5(4), 708–711 — a critical review by a leading measurement psychologist questioning the validity of the IQ test used and the plausibility of the gains.

Rosenthal responded to these critiques over subsequent years and conducted meta-analyses of the broader expectancy literature. The meta-analytic picture was more supportive than any individual replication: a synthesis of hundreds of studies found consistent evidence for interpersonal expectancy effects, though generally smaller than the Pygmalion original suggested.

Source: Raudenbush, S.W. (1984), "Magnitude of teacher expectancy effects on pupil IQ as a function of the credibility of expectancy induction," Journal of Educational Psychology 76(1), 85–97 — a meta-analysis showing that expectancy effects are real but moderated by how plausible the expectation is; effects are larger when teachers have little prior knowledge of the student.

What Meta-Analysis Has Established

By the 2000s, enough research had accumulated to draw more confident conclusions. The core finding — that teacher expectations influence student performance — is now well-supported. A comprehensive meta-analysis by Jussim and Harber in 2005 reviewed the entire expectancy literature and concluded that the effect is real but substantially smaller than the Pygmalion original suggested, and that it is strongest under specific conditions.

Source: 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 — the definitive meta-analytic review of the teacher expectancy literature.

The Jussim and Harber review found three conditions under which expectancy effects are largest:

1. Low prior information: When teachers have little basis for forming an independent judgment of a student's ability, externally provided expectations carry more weight. The Pygmalion effect is largest for students new to a teacher, at the start of a school year, or in early grades before teachers have formed their own assessments.

2. Stigmatized groups: Students from groups that face negative stereotypes show larger expectancy effects in both directions. Positive false expectations produce larger gains; negative false expectations produce larger deficits.

3. Younger students: Younger children are more dependent on teacher interaction for their cognitive development and less able to compensate for low teacher investment through independent effort.

Importantly, the Jussim and Harber review also found that most teacher expectations are accurate rather than distorting — teachers generally form reasonably correct assessments of their students and these accurate expectations are not "effects" in the causal sense being studied. The Pygmalion effect is about what happens when teachers hold beliefs that deviate from the student's actual ability, and in real classrooms, genuinely false beliefs are less common than the experiment created them to be.

The Replication Record: Where Expectancy Effects Hold and Where They Fade

The Pygmalion original reported double-digit IQ gains in some grades, which attracted excitement and skepticism in roughly equal measure. What followed was four decades of replication attempts, meta-analyses, and theoretical refinement that produced a picture more nuanced — and ultimately more credible — than either early enthusiasts or early critics were willing to concede.

The most important early synthesis came from Stephen Raudenbush, then at Michigan State University, who in 1984 published a meta-analysis of 18 studies that had experimentally induced teacher expectations and measured subsequent student IQ or achievement. His central finding reframed the entire debate: the magnitude of the expectancy effect depended strongly on how much prior knowledge teachers already had of their students when the false expectation was introduced.

Source: Raudenbush, S.W. (1984), "Magnitude of teacher expectancy effects on pupil IQ as a function of the credibility of expectancy induction," Journal of Educational Psychology 76(1), 85–97 — a meta-analysis of 18 experimental studies showing that effect size is moderated by teachers' prior familiarity with students.

Prior familiarity — the degree to which a teacher has already formed an independent assessment of a student before receiving an externally provided expectation — is the key moderator. When expectations were induced before teachers had met their students (as in some studies that delivered the bloomer list before the school year began), effects were meaningful. When expectations were induced after teachers had already spent weeks or months with students and had formed their own judgments, effects were small to negligible. A teacher who already knows a student is performing at grade level cannot easily be persuaded that the student is a hidden genius — her own classroom observations carry more evidential weight than a test result from an external researcher. Milgram's design worked partly because the expectation was planted at the very start of the school year, before teachers had accumulated contradicting evidence.

Raudenbush's synthesis helped explain why replications were so inconsistent: studies that closely matched the Pygmalion timing (expectation introduced before or at the very start of teacher-student contact) tended to find effects; studies that introduced the manipulation later in the year tended not to. The original Rosenthal-Jacobson design had inadvertently optimized for the condition under which the effect would be largest.

By the mid-2000s, the literature had grown large enough for a more ambitious synthesis. Lee Jussim and Kent Harber at Rutgers University reviewed the full body of teacher expectancy research and published what remains the field's reference review. Their conclusions were deliberately measured and, to many education researchers, somewhat deflating.

Source: 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 — the most comprehensive review of the expectancy effect literature, covering experimental, correlational, and naturalistic studies.

Jussim and Harber confirmed that expectancy effects are real. They are also, in most naturalistic classroom settings, small. The typical effect size found in correlational studies — where researcher-implanted false beliefs are absent and teachers simply form their own judgments — accounts for roughly 5% to 10% of the variance in student outcomes. That is not zero, but it is far from the 10–12 IQ-point swings the Pygmalion original reported. The review also drew a distinction that the popular reception of the experiment had largely collapsed: most teacher expectations are accurate. Teachers generally read their students' ability levels with reasonable fidelity. The expectancy effect studied by Rosenthal and Jacobson is specifically about what happens when expectations are inaccurate — and in normal classrooms, genuinely false beliefs are rarer than an experiment requiring deception can easily create.

The review identified the conditions under which effects are largest and most likely to accumulate across time:

• Stigmatized groups: Students from groups subject to negative stereotypes (defined by race, socioeconomic status, or perceived disability) are more vulnerable to inaccurate downward expectations and show larger effects in both the positive and negative direction.
• Early grades: Younger children are more dependent on teacher input and less able to self-correct through independent study or peer comparison.
• Low prior information: Consistent with Raudenbush, effects are larger when teachers lack independent evidence to anchor their judgments.
• Accumulation across years: Effects that are small in a single year may compound if a student consistently encounters teachers who share the same inaccurate prior. A student tracked into a low-ability group at age seven carries the expectation signal across multiple teacher relationships.

Self-fulfilling prophecy — distinct from accurate perception — occurs specifically in the compounding case: a false belief shapes behavior, behavior shapes outcome, and the outcome then provides apparent confirmation of the original belief, making correction increasingly difficult over time.

The modern consensus is neither the triumphant reading that expectations rewire intelligence nor the dismissive reading that the Pygmalion effect is a statistical artifact. Teacher expectations function as one real input among many into student development, with effects strongest at the moments when the teacher-student relationship is newest and the student's identity as a learner is most malleable. That is a narrower claim than Rosenthal and Jacobson made in 1968. It is also a claim that has survived decades of scrutiny.

The Broader Implications

The Pygmalion effect has traveled far from education. In medicine, research has shown that patient outcomes improve when physicians hold positive expectations about recovery and communicate those expectations to patients — related to but distinct from placebo effects. In management, studies have found that supervisors who hold high expectations of new employees produce better performance, largely through the same CATE channels: more coaching, more complex assignments, more patient feedback.

Source: Eden, D. (1990), "Pygmalion in management: Productivity as a self-fulfilling prophecy," Lexington Books — the extension of expectancy effects to organizational and management contexts, including field experiments in military units.

The implications for education policy remain contested but consequential. If teacher expectations influence student outcomes independently of student ability, then the mechanism by which low-income and minority students underperform in tracked or ability-grouped classrooms may partly reflect the self-fulfilling expectations created by the tracking decision itself. Placing a student in a low-ability track communicates an expectation of low performance to teachers, who then teach that student differently, confirming the original assessment.

This does not mean that ability grouping always produces expectancy effects, or that expectancy effects are large enough to account for observed achievement gaps. But it identifies a channel through which institutional decisions can shape individual outcomes in ways that are difficult to detect or correct.

The Takeaway

The Oak School experiment was methodologically imperfect, and the original effect size — double-digit IQ gains in some grades — has not held up in the replication literature. But the core finding is real: teacher expectations, even when planted falsely, produce systematic differences in how teachers behave, and those behavioral differences produce measurable differences in student outcomes. The effect is largest early, largest when teachers have little prior information, and largest for students from stigmatized groups who are most vulnerable to having their apparent ability misread. Decades of research have built this from a contested single study into one of the most replicated phenomena in educational psychology. The practical lesson is not that all expectations become self-fulfilling, but that expectations are not passive predictions — they are active forces in the environment they are predicting.

---

Read next:

• The Asch Conformity Experiment
• The Rosenhan Experiment
• The Bobo Doll Experiment

---

Q: What was the "Harvard Test of Inflected Acquisition" used in the experiment?
A: It was a fictional name invented by Rosenthal and Jacobson to make an ordinary IQ test sound like a specialized developmental screening tool. Teachers were told it could predict which students were about to enter a period of rapid intellectual growth. No such test existed. The actual instrument administered was the Flanagan Tests of General Ability, a standard IQ battery.

Q: Were the IQ gains in the Pygmalion experiment real measured improvements or could they be testing artifacts?
A: The gains were measured on a standardized test given by the same researchers at baseline and follow-up, so they represent real score changes. Whether those score changes reflect genuine improvements in cognitive ability or more superficial test-familiarity effects is harder to determine. Critics like Thorndike questioned whether the IQ gains, especially the very large first-grade gains, were within the plausible range for a real developmental effect or more likely reflected administration or scoring irregularities.

Q: How do expectancy effects reach children — do teachers consciously treat students differently?
A: The evidence suggests that most expectancy-driven behavior is not conscious. Teachers generally do not decide to treat "bloomers" better — the behavioral differences in warmth, wait time, feedback specificity, and material difficulty appear to operate outside conscious awareness, through the same subtle channels that shape any interpersonal interaction where one person holds expectations about another.

Q: Does the Pygmalion effect work in reverse — do negative expectations harm students?
A: Yes. The same meta-analyses that document the positive expectancy effect find a corresponding negative effect when teachers hold falsely low expectations. The negative effect tends to be smaller in magnitude than the positive effect, possibly because students have other sources of information about their own ability that counteract teacher pessimism. But in stigmatized groups and younger children, the negative effect is larger and more persistent.

Q: What happened to the Oak School students identified as "bloomers"?
A: Rosenthal and Jacobson followed up at the end of the second year (1966). The IQ gains for first- and second-graders largely persisted. For older grades, gains diminished, suggesting that the critical window for expectancy effects may be narrowest in the early years of school. Long-term follow-up beyond two years was not conducted.

Q: Is the Pygmalion effect the same as the placebo effect?
A: They share a structure — a belief influences an outcome — but the mechanisms differ. The placebo effect primarily operates through the patient's own physiological and psychological response to a perceived treatment. The Pygmalion effect operates through a third party's behavior: the teacher changes how they interact with the student, and it is those behavioral changes, not the student's direct response to the expectation itself, that drive the outcome.