Abstract The heads of submarine canyons represent a critical link in the transfer of sediment from terrestrial sources to deep basin sinks. Here we report data on grain size, bathymetry, and geochronology from twenty-five modern submarine canyons that suggest this link to be very sensitive to the distance between the canyon head and the shoreline, and, to a lesser extent, wave energy. These data show the width of this zone filters the caliber of sediment delivered into deep water, which has significant implications for understanding sediment budgets and the distribution of reservoir and seal facies. Data from modern systems show that the river mouths or longshore drift cells must come within about 500 m of the head of the canyon to deliver gravel-size material and within 1 to 5 km to deliver sand-size material to be transported down the canyon into deep water. Clay- and silt-size particles are transported greater distances across the shelf, up to a few tens of km, whereas beyond about 40 km, little sediment makes the connection to the heads of canyons and deposits are dominated by condensed, carbonate-rich sediments. Our data from modern systems are consistent with existing sequence stratigraphic models for sediment delivery to deep water. The significance of our work is to show in more detail how and when connections can occur between fluvial to shallow-water systems and submarine canyons and how these connections regulate the quantity and caliber of sediment that can be transported into deep water. Once the process-based conditions for connection are met, then the geology and climate of the source area control the quantity and caliber of sediment that can be moved to deep water. We hypothesize that connection times, and the resultant fractionation of sediment mass and grain size between shelf and deep-water depocenters, may have varied in a predictable way through geologic history. For example, during greenhouse times when sea level was relatively high, but with inherently low high-frequency variability, longer-lived connections between fluvial to nearshore environments and deep water may have been more likely. This scenario would favor the preferential transfer of sediment, especially sand, into deep water, and the development of thick, laterally extensive sand-rich basin-floor deposits. By contrast, during icehouse periods, high-amplitude sea-level fluctuations and inherently wider continental shelves may have resulted in repeated landward and seaward transits of river mouths and shorelines, shorter connection times between source and sink, especially for sand-size sediment, and preferential sequestration of sediment in shelf to shelf-margin parts of the system. These conditions would have resulted in deep-water deposits that are a mixture of locally thick sands, abundant turbidity-current-derived mud, and thin but basin-wide condensed sections that represent periods of sediment starvation in deep water.