If gels are investigated in an analytical ultracentrifuge by means of sedimentation equilibrium experiments, so called soluble parts which are originally not incorporated into the network influence the swelling pressure equilibria of the gels. For gelatidwater, it can be shown that an irreversible process takes place which is dependent on the rotational speed of the experiment. Experimental evidence is given that soluble parts which are partly associated and aggregated to the polymer network can act as a branching and/or crosslinking agent. A model for this process is given. In such a case an application of the results of the swelling theory, e. g. that of Flory and Huggins assuming a constant number of network junctions, cannot be applied anymore. A universal procedure is described how swelling pressure equilibria can be identified, although the crosslinking density is changed. It can be shown that the gradient of the soluble parts, respectively the additional structural changes of the gel by the soluble parts, is reproducible and with it the swelling pressure equilibria. Furthermore the radial distribution of dyed soluble parts inside the gel phase is presented as a function of time during a sedimentation equilibrium run with a gelatidwater gel. It can be shown that the back diffusion process of the soluble parts in the gel is very slow and does not lead to the sedimentation-diffusion equlibrium of free soluble parts in the considered time interval of 4 months. In combination with the detected fast sedimentation of the soluble component in the gel phase this gives further evidence for the association of the soluble parts to the gel network forming a gradient gel. The predicted formation of the anisotropic gel could be proved by a swelling experiment. a).