Acoustic properties determined from logs and discrete measurements along a borehole transect across the slope of the western Great Bahama Bank margin (ODP Leg 166) show a distinct relationship with primary sediment composition and noncarbonate content. Dominant sediment types are alternating cemented and uncompacted calciturbidites with shallow-water derived grains, and poorly to uncemented and compacted periplatform ooze of globigerinid-dominated wackestone to packstone. Comparison of the velocity-porosity variation from logs and discrete measurements reveals that both tend to follow the trend of generally accepted velocity transforms. Primary control on the acoustic velocity is exerted by total porosity, whereas the remaining variation in velocity is explained by differential diagenesis. The less cemented to uncemented and compacted background sediments have velocities that, at a given porosity, fall below the time- average equation for calcite. The cemented turbidites, however, have velocities that are higher than those predicted by the time-average equation. Although it is unclear whether the original volume of aragonite or presence of organic matter controlled this difference, it is suggested that the acoustic properties reflect an input signal. The presence of noncarbonate material, some of which is clay, increases basinward and further reduces the acoustic velocities in the background sediment. In contrast, earlier work on the more proximal slope, and nearly pure carbonate, intervals in Clino and Unda cores documents diagenetic differentiation in compositionally similar sediment, resulting in alternating acoustically fast and slow layers. Marginal and nearly pure limestones have a more complex acoustic behavior and hence are difficult to predict. Though both velocity logs and discrete velocity measurements show comparable general trends, logging data show a considerably larger variation at a given porosity value (up to 3.5 km/s) than discrete data, which ranges between 1.0 and 2.5 km/s. To improve the quality of velocity logs it is suggested to groundtruth these data with other logging data, general velocity transforms, and, ideally, discrete measurements on core plugs.