How the processes that initiate turbidity currents influence turbidite deposition is poorly understood, and many discussions in the literature rely on concepts that are overly simplistic. Marine geological studies provide information on the initiation and flow path of turbidity currents, including their response to gradient. In case studies of late Quaternary turbidites on the eastern Canadian and western U.S. margins, initiation processes are inferred either from real-time data for historical flows or indirectly from the age and contemporary paleogeography, erosional features, and depositional record. Three major types of initiation process are recognized: transformation of failed sediment, hyperpycnal flow from rivers or ice margins, and resuspension of sediment near the shelf edge by oceanographic processes. Many high-concentration flows result from hyperpycnal supply of hyperconcentrated bedload, or liquefaction failure of coarse-grained sediment, and most tend to deposit in slope conduits and on gradients < 0.5° at the base of slope and on the mid fan. Highly turbulent flows, from transformation of retrogressive failures and from ignitive flows that are triggered by oceanographic processes, tend to cannibalize these more proximal sediments and redeposit them on lower gradients on the basin plain. Such conduit flushing provides most of the sediment in large turbidites. Initiation mechanism exerts a strong control on the duration of turbidity flows. In most basins, there is a complex feedback between different types of turbidity-current initiation, the transformation of the flows, and the associated slope morphology. As a result, there is no simple relationship between initiating process and type of deposit.