In the present investigation the velocity sedimentation technique was analysed with respect to separation resolution, power and sensitivity. It was found that apparative modifications do not influence the resolution, which is a function of the contribution of apparative errors to the dispersion. A surprisingly small parameter of 0.15 was determined and it seems unlikely that this value can be improved. On the other hand an apparative modification is presented which improves the separation power and makes sample loading independent of the gradient filling. If cells (from rat bone marrow) were separated, a several times higher dispersion for a given cell volume was observed than was due to the apparative error. It was concluded that density variations were the major source of this dispersion. Since cell volume and density apparently show independent variations within a biological cell population the cell density cannot be disregarded if velocity sedimentation profiles are discussed in physical terms as is often done.