Recent literature on interpolating retrieval practice compared two learning « grain sizes » a fine grain, and a larger grain with a certain variety of learning contents. There should be an optimal grain size for the retrieval practice condition, as well as for the re-read condition ; and there should be a optimal grain size for which there is the strongest testing effect. In the present study, we cross the retrieval effect and the grain size effect at different retention intervals. First, we aimed at comparing the effect of different grain sizes of learning periods on memory retention. Second, we wanted to measure for which grain size we will find the strongest testing effect.  The experiment was entirely run online on the Didask digital learning platform (https://www.didask.com/). Study material consisted of a course on cognitive balance at work in light of cognitive sciences provided by the company Cog’X. We used a mixed factorial design that included 2 between-subject Learning Conditions (quiz-reading, reading-reading) and 3 between-subject Grain Sizes (small, medium, large) for the acquisition phase. For the final test we manipulated 2 within-subject Question Types (trained, untrained) and 2 within-subject Retention intervals (7 days, 28 days). We recruited more than 300 adult participants via the French online work platform Foule Factory. During the training phase at day 1, participants had to learn from contents according to the learning conditions to which they were assigned. Seven days later, they had to complete the final exam for the first time, and they had to do again 21 days later. The experiment is not finished, we are currently collecting the data.