Five parental pairs of barley (Hordeum vulgare L.) were selected on the basis of wide differences in heads/30 cm of row (X), kernels/head (Y), and mg/kernel (Z). The crosses and backcrosses to each parent were allowed to self to the equivalent of the F4 at which time 20 random selections were drawn from each population. Seed was increased for two generations and standard nursery plot trials were conducted on the equivalent of the F7. The objectives were to observe component complementation and to see how it affected predicted yield.We were able to predict progeny mean yields (W) the basis of parental (XY). In only one case was there a significant failure in prediction. The backcross 1 ✕ 42 significantly exceeded the regression estimate and the mean of the higher parent ∗ and this for the selfed unselected progeny. Comparisons via a graph of Y on X showed that parent 4 was an ‘outlier’ and that it somehow relaxed the constraints in X on Y permitting a large (XY) and higher yield which was expressed in the progeny.All parents had an acceptable level of W but differed widely in component values. Mated pairs were selected to complement component values so that these were phenotypically wide crosses. In every case the mean yield of one or the other backcross exceeded that of the cross. Using vector terminology, the backcross narrowed the angle between parents. In terms of developmental physiology and morphology, the backcross tended to preserve the integrity of an already proven system while adding bits and pieces of another system. When the parents are more similar, the advantage of the backcross would be expected to dissipate.