Sediment acoustic models contain two connected components, the geo-physical description of the sediment and the model of acoustic processes. Geo-physical descriptors are used in the classification of sediments, and they are based on grain size, density and other physical descriptors. Acoustic sediment models were initially fluid approximations that were simple to implement. As the need for accuracy increased, the fluid model was extended to stratified fluid and visco-elastic models. The latter, with five frequency-independent parameters, appeared to be consistent with sediment acoustic data up to the 1980s. More recent experimental data have revealed discrepancies in the frequency-dependence of attenuation and sound speed, particularly in the case of sandy sediments, which cover a large fraction of the continental shelves. Broad-band acoustic measurements of wave speeds and attenuations are more consistent with a poro-elastic model, consisting of Biot's theory with extensions to account for the physics of granular media. Aside from terminology, there is a fundamental difference between viscoelastic and poro-elastic theories. The former is based on two types of waves, a compressional wave and a shear wave, while the latter has an additional compressional wave, often called the Biot wave. There are currently two approaches to the development of sediment acoustic models: (a) visco-elastic models with frequency dependent parameters that mimic the observed behavior, and (b) poro-elastic models that reflect the physical processes. It is shown that (a) would be a significant improvement over existing models, but (b) is the preferred solution.