This article reports on a pilot study on the academic self-efficacy (ASE) of 245 Finnish pupils when studying according to a subject-specific curriculum for one week and an integrated curriculum the following week. The data consisted of selected items from the Class Maps Survey and self-reported grades of L1 and mathematics and were analysed using latent growth curve modelling and multivariate linear regression analysis. According to the results, the mathematics grades were associated with the level of and change in ASE: pupils reporting lower skills had lower self-efficacy at the beginning of the study. During the first week, ASE increased similarly among all skills groups; during the second week, only the math grades positively predicted variation across all pupils in rate of change in ASE. The increase in ASE during the second week was faster among pupils with low or good math grades than the pupils with higher math grades.