VI
Mark học AIPygmalion in the Classroom · Robert Rosenthal and Lenore Jacobson, 1968
If a teacher believes a pupil is about to improve, does the pupil really improve? Today's answer: a little, mostly when the teacher knows nothing else about the child, and far less than the legend of "believe and they become geniuses".
Teachers were "fooled" about one detail only: the list of bloomers was actually random.
Harris and Rosenthal (1985) pooled studies of these behaviours: expectations do not travel by telepathy, but through small gestures repeated every day. Climate and input are the clearest channels.
Fooling a boss with a fake list works; teaching a boss to "expect more" is much harder. That is why the lessons in Part 6 focus on concrete behaviour, not belief.
10 common claims about the Pygmalion effect, checked against current research. Click each line for the reason.
Rosenthal and Rubin (1978) pooled 345 experiments on interpersonal expectancy effects. Even the strongest critics (Jussim and Harber, 2005) conclude the effect is real. The debate is about size, not existence.
Harris and Rosenthal (1985) pooled studies on how people with expectations behave: warmer, more teaching, more chances to answer. The mechanism is fairly consistent.
Expectations correlate with achievement at about 0.4–0.8, while experimental effects are about 0.1–0.2. Jussim and Harber (2005) estimate about 75% accuracy, about 25% self-fulfilling prophecy.
The overall gap was only about 4 IQ points (12 vs 8), almost all from grade 1 (about 15 points) and grade 2 (about 10 points); other grades showed no difference or the reverse (per Jussim and Harber, 2005). Thorndike (1968) pointed out that one whole class had a mean pre-test IQ in the intellectually disabled range. Snow (1995) found about 35% of scores outside the test's norm range, and the effect vanished without 5 extreme pupils (one went from 17 to 110).
Madon, Jussim and Eccles (1997): effects of about 0.26 for low achievers, 0.08 for high achievers. Jussim and Harber (2005) report similar findings for poor and African American pupils, but only about 2 studies have tested this.
McNatt (2000) pooled 17 studies (2,874 people): mean d of 1.13. But most were training courses, many in the military, and effects varied widely. There is little data from ordinary, running companies.
Yeager et al. (2014), US 7th graders: among African American pupils, 64% turned in a revision with this note vs 27% with standard feedback. Clear result, but in a school, small sample, little tested at work.
Babad, Inbar and Rosenthal (1982): bias-prone student teachers treated pupils they saw as weak more negatively, and the gap showed in task performance. Only 26 student teachers, but consistent with later studies.
Raudenbush (1984) pooled 18 IQ experiments: mean d = 0.11; Snow (1995) put the median at just 0.035. The effect is near zero once teachers have known pupils for more than 2 weeks. The "believe and they become geniuses" version in self-help books is an exaggeration.
Eden et al. (2000) ran 7 experiments training managers in a "Pygmalion leadership style": small mean effect (r = 0.13) and little evidence of change in bosses or staff. Eden, the pioneer of workplace Pygmalion research, published this himself.
A real but small effect, strongest when people are still strangers. Keep the lesson about behaviour, drop the promise of magic.
Pygmalion arrived just as the US was fiercely debating poor and Black pupils' education. The story "if teachers believe, children will thrive" was hopeful and simple, so it spread through the press, education books and later management books.
Thorndike (1968), a measurement specialist, wrote a scathing review: some IQ data could not possibly be right, so every conclusion drawn from them was doubtful. His famous line, paraphrased: when the clock strikes thirteen, you doubt more than the last stroke (paraphrase, not an exact quote). Rosenthal replied that poor measurement makes differences harder to find, so the differences found still mean something.
Dozens of later experiments give a modest picture: the effect is real, usually small, larger for disadvantaged people, and most of the predictive power of expectations comes from people guessing right. It is an early example of what is now called the replication crisis.
You do not need to "believe harder". You need to check how you treat people differently across the 4 channels, then even it out.
Every number in the article, the Facebook slides and captions comes from the studies below.