Teaching experiments with pairs of children have generated several hypotheses about students' construction of fractions. For example, Steffe (2004) hypothesized that robust conceptions of improper fractions depends on the development of a splitting operation. Results from teaching experiments that rely on scheme theory and Steffe's hierarchy of fraction schemes imply additional hypotheses, such as the idea that the schemes do indeed form a hierarchy. Our study constitutes the first attempt to test these hypotheses and substantiate Steffe's claims using quantitative methods. We analyze data from 84 students' performances on written tests, in order to measure students' development of the splitting operation and construction of fraction schemes. Our findings align with many of the hypotheses implied by teaching experiments and, additionally, suggest that students' construction of a partitive fraction scheme facilitates the development of splitting. (Contains 8 tables and 6 figures.)