The brain forms folded shapes to compartmentalise the complex neural circuitry. How the amount of folding is determined during development is not well understood, as most modelling-based predictions are derived from adult brains. Here, Andrew Lawton and colleagues test multiple predictions of folding by comparing two mouse strains with naturally distinct amounts of folding. First, they identify a postnatal period when the folding diverges in the two strains and observe regionally distinct levels of folding in individual lobules. During this period, the external granule layer (EGL) and the underlying cerebellar core expand at different rates, and the level of EGL/core differential expansion determines the final folding in different regions. The authors also observe that the EGL thickness correlates with the final folding wavelength — the distance between the base of each pair of fissures. Furthermore, they find that the regional variation in Purkinje cell densities in the two strains can account for the folding differences. Finally, they observe that cell division angle in the EGL can predict the level of differential expansion and the EGL thickness. Overall, the findings suggest that regional variation in cell division angle can set the amount of folding in the brain by regulating the level of EGL/core differential expansion and the EGL thickness.