Infants Possess a System of Numerical Knowledge

which the manipulation of teacher expectations was quite plausible be cause teachers had known pupils only for 2 weeks or less, 9 studies show positive effects (i.e., 90%, compared with Snow's 56%). Simi larly, Snow reports the median effect size (cf) for all 18 studies as .035, or .025 omitting the Pygmalion study. For the 10 studies of the more plau sibly manipulated expectancies, the median effect size was .195, or .18 omitting the Pygmalion study. In an updated analysis of 19 studies of teacher expectancy effects,5 Rau denbush estimates that the expected effect size (d) for studies having no prior teacher-pupil contact is .43, a substantial effect 3 to 6 times larger than the effects of aspirin described in my article6 on which Snow is commenting. Snow refers to the original Pyg malion study7 as "discredited." But that should no longer be the issue. Even if Lenore Jacobson and I had never conducted our experiment, there are now too many new studies for even committed criticisms of dis liked results to make the basic con clusion go away: Teachers' expecta tions can affect pupils' intellectual functioning. Science is the loser when new data have no effect on prior belief.