I've created a task, assigned to myself, to enable hybrid models in relax. The task is located at https://gna.org/task/?func=detailitem&item_id=3122. An example of the use of a hybrid model could be if the protein consists of two independent domains. These two domains could be analysed separately, each having their own optimised diffusion tensors. The N-terminal domain run could be called 'N_sphere' while the C-terminal domain could be called 'C_ellipsoid'. These two runs could then be hybridised into a run called 'mixed model' by using the hypothetical user function 'run.hybridise()' which can then be compared, via model selection, to a run where the entire protein is assumed to have a single diffusion tensor. The only requirements for runs to be hybridised is that, at minimum, a sequence has been loaded, that the sequence for all hybridised runs is the same, and that no residue is allowed to be selected in two or more runs. The last condition is to ensure that overlap does not occur to allow statistically significant comparisons. Edward