The most widely used, although not as widely documented, explanation for initiation of turbidites is the conversion of catastrophic sediment failures into turbidites. Such instantaneous sediment failures generate short-lived surge-type flows that deposit turbidite beds described as Bouma sequences. In this communication, we present data from the Eocene Central Basin of Spitsbergen that shows turbidite systems that are volumetrically dominated by thick ungraded and laminated sandy turbidite beds, deposited by gradual aggradation from sustained flows. Sustained flows are long-lived and more or less continuous. They can be generated by a variety of mechanisms, e.g.: (1) instability during volcanic eruptions and the consequent remobilisation of unconsolidated material; (2) seismically triggered subaerial sliding within the drainage; (3) storm surges; (4) retrogressive slope failure; (5) breaching; and (6) hyperpycnal flows. Our database shows that in the Eocene Central Basin, the sustained flows were generated by hyperpycnal flows, i.e., by direct river effluent. The hyperpycnal flow turbidites have been recognized on the basis of (1) physical connection between fluvial and turbidite channels at the shelf edge, (2) abundance of thick turbidite sandstone beds, (3) sand-prone nature of the turbidite systems, (4) downslope changes of individual thick, sandy turbidite beds, and their collapsed pinch-out segments, (5) great abundance of continental material (leaves, coal fragments) in turbidite beds, (6) low abundance of associated slumped or debris-flow beds, and (7) the occurrence of turbidites in systematically accreted shelf-margins. These features strongly suggest that hyperpycnal flow, generated by direct river effluent, deposited significant number of turbidite beds in the documented slope and basin-floor turbidite successions. We also argue that the hyperpycnal flows are preferentially fed into the deepwater slopes beyond the shelf edge during the falling stage and lowstand of relative sea level. The turbidite beds were studied in the context of seismic-scale (1×15 km), dip-oriented outcrops by measuring detailed vertical sections, following individual beds and packages of beds downslope for distances over 5 km, lateral mapping, and helicopter-taken photomosaics.