A winter barley population composed of 387 lines in the F5 generation was grown in 1966 from which 136 strains were selected with a mean yield of 123.3% of the cheek variety. In the following generation the same 136 selected strains were grown in 1967 producing a mean yield of 90.7% of the same check variety.The “apparent” negative response to selection was analyzed with respect to yield components: heads per unit area (X), kernels per head (Y), and kernel weight (Z). Twice as many heads were produced in 1967 as 1966, but fewer kernels per head and smaller kernels were produced in 1967 than in 1966. The inverse relationship between the components is explained by “component compensation.”The inter‐annual correlations for the yield components were slightly negative. The overriding effect of the environment on the genetic processes and extreme compensation between yield components were postulated to cause the low inter‐annual correlations.Evidence was presented that optima exist in any given gene pool for yield components and their inter‐relationship in order that maximumy ields can be reached for a given environment. Selection in one environment for these optima does not mean necessarily that under another environment the same production performance will be obtained. This in turn suggests that the plant breeder should consider the joint manipulation of genotype and environment to get the maximumg ain from a given gene pool. He must either do this or accept compromises which fail to achieve maximum results.