Of numerous theories of drumlin genesis, none has been widely accepted. It seems evident, however, that some form of positive feedback process is involved. Under certain circumstances perturbations are amplified. Herein we suggest that patchy areas of frozen bed provide the initial perturbation. Such frozen patches may occur in local areas underlain by material of lower thermal conductivity or on slight topographic highs. Drag exerted by the frozen patch deflects ice flow into its lee, dragging with it mobile till eroded from the thawed area. The energy balance is such that this till likely refreezes, either producing a topographic perturbation or amplifying an existing one. The resulting topography then deflects more of the geothermal heat away from the developing hill and into the adjacent trough, resulting in a positive feedback. Once the thermal perturbation exceeds a critical (though as yet undefined) level, melting may decouple the ice from the bed, preventing further entrainment of till from thawed areas, and thus limiting the height and length of the drumlin.