The feeding preferences and comparative fecundity of laboratory‐bred, alienicolae alatae ofAphis fabaewere investigated in small leaf cages on spindle and sugar beet leaves, representing a winter and a summer host respectively, both in pots in the greenhouse and while growing naturally outdoors.The aphids' readiness to stay and feed, and their average reproduction rate, both varied between leaves of different ages and kinds, and there was evidence of true fecundity differences among aphids feeding on different leaves. As a general rule, readiness to settle and reproduction rate varied together in the same sense, but they did so significantly less often in comparisons between leaves of different kinds, spindle and beet, than in comparisons between leaves of one kind only. Among the leaves on the same kind of plant, the aphids preferred to feed and reproduced faster, on the whole, on young and early senescent leaves than on mature ones. To a limited extent, this rule seemed to govern also the aphids' comparative readiness to settle and reproduce, as between leaves of the two different kinds of plant. Allowing for age‐differences among the leaves, the aphids settled and reproduced better on spindle than on beet leaves. Since the aphids used were alienicolae, this tends to confirm the usual designation of the winter host as primary and the summer hosts as secondary, for a given aphid species.These findings form the basis of a dual discrimination theory of host selection, which assumes that aphids respond behaviourally to at least two main classes of leaf property: one associated with the age of the leaf and the other with the kind of plant. It is suggested that in nature these two sensory requirements of the aphids may be essentially contradictory, and that the shifting patterns of aphid distribution among leaves and plants may depend on the shifting distribution of leaves offering a satisfactory compromise between them. The phenomenon of host alternation is considered in the light of the dual discrimination theory, as a particular instance of a shifting distribution‐pattern probably connected with the alternation of the seasons of active growth and senescence in the winter and summer host p