The phenomenon of experimenter bias (E-bias) was investigated in a fourth-grade arithmetic lesson by using a modified programmed learning task on division in the five-family (five is one of the factors.) E-bias predicts a relation between teacher expectancies and outcomes despite standardization of tasks. Six schools were randomly assigned across three treatments: positive, negative, and neutral bias. Two teachers from each school were randomly selected. The two-stage nested ANOVA was not significant (p<.20), although the treatment means were ordered positive bias, neutral bias, negative bias (high to low.)