Existing hypotheses on the evolution of feathers are reviewed with the assumptions that feather evolved from reptilian scales and that pennaceous feathers evolved before downy feathers. Observations with a scanning electron microscope demonstrate that basic to the structure of pennaceous feathers is the lamelliform structure of barbules, the planes of which are oriented at right angles to the plane of the feather vane. Thus the structure of the vane is more open than generally realized. The airtight vane of flight feathers is assumed a later specialization. Most of the existing hypotheses assume that the feather acts as a relatively solid barrier between the skin of the bird and the exterior and they are therefore not in agreement with the actual structure of feathers. A hypothesis is needed which explains the adaptive value of a pennaceous feather being porous. The hypothesis is put foward that feathers evolved due to selection for a water-repellent integument. For purely physical reasons a porous surface repels water drops more strongly than does a solid surface of the same material. Physicists have pointed out that the structure of feathers conforms closely with the theoretical requirements for water-repellency. Possibly feathers started to evolve on reptiles living at the seashore, where the main advantage of increased water-repellency was to reduce cooling from evaporation of water off a wet integument.