Kelp gulls peck at a red spot on their mother's beak. Geese and cranes imprint on hang gliders. Presumably, these processes depend on sensory primitives (red spots, triangular shapes), and these are not learned (via e.g. reinforcement learning), but genetically set. Are there any current ideas of how (not why) these abstract, high-level sensory features are genetically encoded?
Naively, some pre-wired relationship between certain retinal neurone ("red"), higher neurons responsible for shape detection ("circular") and association cortices regulating a pre-specified behaviour is assumed.