How can it be explained that, aside from inter-individual differences, pupils in certain classes are more responsive than others to the formal aspects of a problem that has been set? The author puts forward the hypothesis that teachers differ in their ability to operate relevant variations in the conception of problems. The differences in inter-class variability could explain the differences in the responsiveness of pupils to the implicates mobilised in the didactic contract. The author examines 42 addition and subtraction problems produced by 7 teachers, using a model for measuring variability, made up of 14 variables; the results obtained do not invalidate his hypothesis: he records a close correlation between the measurement of responsiveness and that of variability. (Contains 5 tables, 3 figures, and 13 notes.)