This longitudinal cross-lagged panel study investigated the development of the structure of young children's spontaneous number focusing tendencies and their longitudinal associations with numerical abilities and mathematics achievement over the course of a 3-year period, that is, in the second and third year of kindergarten and in first grade of primary school in Flanders (Belgium). Participants included 177 children (50% boys) with a mean age of 4.35 years at the start of the study. Each school year, children completed measures of their spontaneous focusing on Arabic number symbols (SFONS), spontaneous focusing on numerosity (SFON), numerical abilities, and mathematics achievement. Confirmatory factor analyses demonstrated that SFONS and SFON were two separate, yet related, constructs across development. Cross-lagged panel models revealed that SFONS and SFON were moderately and differentially related to numerical abilities and mathematics achievement across development. Theoretical, methodological, and educational implications of these findings are discussed.