AbstractFollowing GRIZZLE (1965) and JESDINSKY et al. (1981) it is suggested to replace the traditional two‐groups (AB, BA) crossover design by a four‐groups (AA, AB, BA, BB) design in order to control for treatment‐period interactions (carry‐over effects). Contrary to the above authors it is proposed to use the incrementsD=X2‐X1from 1 st to 2 nd observations rather than the observations themselves as the dependent variable. Testing for main‐ and interaction effects of treatment is treated for increment modified Grizzle‐ and Jesdinsky ANOVAmodels. Alternative and conditional tests are discussed. Numerical evaluations are illustrated by an example from psych