Submarine canyons have generally been lumped together and a common explanation of their origin has been sought. The author suggests that some, named submarine ravines, are drowned river valleys (Corsica). The other extreme, the New England type, is the result of glacial turbidity currents. Daly's explanation of the California type as drowned subaerial valleys, smothered by sediment, and excavated by glacial and recent turbidity currents is accepted with some change of emphasis. This type is thus tentatively ascribed to the same turbidity-current mechanism, but erosion has exhumed some buried valleys, cleaned out parts of others, laid bare parts of old mountain slopes, and has even developed new channels. Shepard's new hypothesis of drowned river valleys kept permanently open by submarine processes, comes near to this picture and is accepted for the submarine ravines, but it cannot account for the features that indicate submarine erosion in the recent past off New England and California. His main argument against erosion by turbidity currents is the absence of scour on lake-delta fronts, but conditions in lakes must be highly adverse to erosion and hence this objection is eliminated. On the other hand Ewing and his associates give convincing evidence for the great importance of turbidity flow in the oceans.