In many practical situations, it is desirable that the students learn all the parts of the material. It is therefore desirable to set up a grading scheme that encourages such learning. We show that the usual scheme of computing the overall grade for the class -- as a weighted average of grades for different assignments and exams -- does not always encourage such learning. Each such intermediate grade describes the student's knowledge of a certain part of the material. From the viewpoint of fuzzy logic, the degree to which the student knows the 1st part of the material and the 2nd part of the material, etc., can be naturally described as a result of applying a t-norm ("and"-operation) to intermediate degrees (intermediate grades) -- e.g., as the minimum of the intermediate grades. It turns out that this fuzzy-motivated min grading scheme indeed encourages students to learn all the material -- and vice versa, the only grading scheme that provides such encouragement is the minimum of the intermediate grades.