The paper describes multivariate analysis of variance in the split plot design when the so-called control treatment B0 is allocated to an additional subplot. The factor B1 allocated to small plots is subject to the two-stage classification, where the levels of the second stage dependent on the levels of the first stage. This dependence is treated as a hierarchical classification of two factors: B1the first stage levels and C (inside B1) the second stage levels.A linear model of the experiment consisting of two forms is used, one form for the plots where all the factors are present and the other for control plots. Beside that, for the sake of better visualisation of the experiment, the component models are based on two different vectors of the grand means. Tests of hypotheses for the comparison with the control are presented. The discussed theory is illustrated by an agrophysical experiment conducted according to such a design.