Measurements of suspended sediment concentration, and the associated bedform morphology, were made beneath regular and irregular waves in a large-scale flume. The bed forms were measured using an acoustic ripple profiler, and the suspended sediments were measured using an acoustic backscatter system, together with pumped sampling. Using the measured bed form dimensions and the flow as inputs, standard approaches have been used to predict the reference concentration and vertical distribution of suspended sediment concentration. Nielsen's <1986> empirical expression for the reference concentration, C0, shows reasonable overall agreement with the present measurements in respect of the dependence of C0 on the Shields parameter (skin friction); but the formula somewhat overestimates the measured concentrations. To analyze the form of the measured mean concentration profiles, comparisons have been made with the simple one-dimensional (vertical) formulations, and extensions thereof, proposed by Nielsen <1992> on the basis of pure diffusion, pure convection, and combined convection and diffusion. It is concluded that in a near-bed layer of thickness about two ripple heights (i.e., the layer dominated by vortex formation and shedding above ripples), pure diffusion characterized by a height-independent sediment diffusivity provides a good representation of the measured profiles. Above this, Nielsen's <1992> convection-diffusion solution provides a better representation. It is shown, however, that by use of pure turbulent diffusion modeling concepts, the same profile can also be obtained by the use of a height-varying, constant + linear, sediment diffusion model. This diffusivity represents the enhanced mixing in the outer part of the oscillatory boundary layer caused by the breakdown of coherent vortex structures into random turbulence. The relative merits of convection and diffusion schemes are discussed.