To simplify the description of selection in two environments the terms ‘ antagonistic’ and ‘synergistic’ are used. Selection upwards in a bad environment or downwards in a good environment is antagonistic, the selection and the environment acting in opposite directions on the character. Synergistic selection is the reverse, upwards in a good environment or downwards in a bad, selection and environment acting in the same direction. Published experiments are reviewed to see how well they agree with two expectations. First, Jinks & Connolly (1973) showed that antagonistic selection reduces environmental sensitivity and synergistic selection increases it. The experiments reviewed showed many exceptions to this rule, but they all showed that sensitivity was less after antagonistic than after synergistic selection. This is shown to be simply the consequence of correlated responses being less than direct responses. Second, I suggested (Falconer, 1989) that antagonistic selection might be the best way to improve the mean performance in the two environments. In the experiments reviewed, antagonistic selection was significantly better than synergistic for changing the mean, but it is now shown that there is no theoretical justification for this expectation; if one type of selection is better in one direction the other ought to be better in the other direction.
Expressions are given for the changes of mean performance and of sensitivity resulting from selection in one or other environment; these changes can be predicted from the parameters of the base population. In the experiments reviewed, an increase of mean performance accounted for 49% or more of the upward response. Equations are presented which allow the variance of mean performance, the variance of sensitivity, and the covariance of mean with sensitivity to be derived from parameters estimated in an unselected population, namely the variances in the two environments and the corresponding covariance. The variance of sensitivity that might be ascribed to scale effects is deduced. Directional selection in a single macro-environment is synergistic with respect to the micro-environmental differences, and is expected to increase environmental sensitivity and consequently to increase environmental variance. Stabilizing selection is antagonistic selection in both directions at the same time, and so is expected to decrease environmental variance.