The "Correlation Length Experiment", an intensive box coring effort on the Eel River shelf (Northern California) in the summer of 1997, endeavored to characterize the lateral variability of near-surface shelf sediments over scales of meters to kilometers. Coring focused on two sites, K60 and S60, separated by ??? 15 km along the 60 m isobath. The sites are near the sand-to-mud transition, although K60 is sandier owing to its proximity to the Eel River mouth. Nearly 140 cores were collected on dip and strike lines with core intervals from < 10m to 1 km. Measurements on each core included bulk density computed from gamma-ray attenuation, porosity converted from resistivity measurements, and surficial grain size. Grain size was also measured over the full depth range within a select subset of cores. X-radiograph images were also examined. Semi-variograms were computed for strike, dip, and down-hole directions at each site. The sand-to-mud transition exerts a strong influence on all measurements: on average, bulk density increases and porosity decreases with regional increases in mean grain size. Analysis of bulk density measurements indicates very strong contrasts in the sediment variability at K60 and S60. No coherent bedding is seen at K60; in the strike direction, horizontal variability is "white" (fully uncorrelated) from the smallest scales examined (a few meters) to the largest (8 km), with a variance equal to that seen within the cores. In contrast, coherent bedding exists at S60 related to the preservation of the 1995 flood deposit. A correlatable structure is found in the strike direction with a decorrelation distance of ??? 800 m, and can be related to long-wavelength undulations in the topography and/or thickness of the flood layer or overburden. We hypothesize that the high degree of bulk density variability at K60 is a result of more intense physical reworking of the seabed in the sandier environment. Without significant averaging, the resistivity-based porosity measurements are only marginally correlated to gamma-ray-bulk density measurements, and are largely independent of mean grain size. Furthermore, porosity displays a high degree of incoherent variability at both sites. Porosity, with a much smaller sample volume than bulk density, may therefore resolve small-scale biogenic variability which is filtered out in the bulk density measurement. ?? 2002 Elsevier Science Ltd. All rights reserved.