Reasons are presented for the view that linkage disequilibrium is of only secondary importance in the general theory of evolution although of primary importance in the theory for particular organisms. It is pointed out that the pattern of factor interaction that is significant in evolution is that which pertains to the mean selective value of populations. The three-phase shifting balance theory of evolution, proposed by the author in 1931, is reviewed briefly as a basis for deciding which sort of interaction system is most pertinent. Some of the misinterpretations of this theory are discussed. It is concluded that the overwhelmingly most important pattern of factor interaction is that in which selection is directed toward an optimum that is never far from the middle of the range of variability. This pattern is characterized by a great many selective peaks. The number rises rapidly with the number of interacting loci and the number of alleles. The effects of linkage in simple two and three factor cases of this optimum model are treated mathematically. The effect is to make the saddles shallower without obliterating them, unless a certain amount of overdominance with respect to selective value is superimposed. The roles of the three phases (random processes within demes as a trigger for intrademic selection toward a new peak, the latter as a trigger for interdemic selection) are illustrated in a hypothetical case with six selective peaks. This indicates a possible interpretation of the 'area effects' of Cain & Currey.