Grebes are unique among birds in ingesting their own feathers. This behavior and the subsequent ejection of feathers as pellets have long puzzled ornithologists, who have tended to treat feather-eating and pellet-casting as independent behaviors rather than as complementary components of the digestive process. The diet of many grebes, including those with the most ancestral traits, is dominated by small invertebrates whose exoskeletons are resistant to digestion. Most birds eat grit to mechanically break down hard foods. Not so with grebes, which are chemical digesters. Feather-eating performs two main functions. The first is to retain food until it is fully digested; this is accomplished by a large feather bolus in the gizzard. The second, provided by a distinct group of feathers in the pyloric pouch, is to filter undigested or indigestible items from entering the intestine. Some of the gizzard bolus is probably regurgitated nightly, but the process is incomplete and undigested food can persist in the gizzard overnight and indigestible hard parts for several months. The pyloric plug is expelled irregularly. Inasmuch as feathers and other debris must eventually be discarded, pellet-casting is an inevitable consequence, not cause, of feather-eating. I propose that grebes originated as surface feeders and adopted feather-eating to enhance the efficiency of feeding on small arthropods or other hard-bodied taxa that are difficult to digest. This interpretation is relevant to understand the early evolution of grebes.