Some properties of the two-period change-over design are investigated. This comparison shows that the two-period changeover design is preferable when the residual effects of the treatments are equal and the correlation between the response to the two treatments is positive. Otherwise the design in which there is random assignment to a single treatment is preferable.