A key challenge when working with capacities is in determining the exponentially increasing number of parameters required for their definition. The issue of tractability is not just with respect to the number of variables, but also in ensuring monotonicity or desirable properties to suit a given application such as submodularity and antibuoyancy. This chapter summarises several approaches that seek a trade-off between computability and the extent to which interaction between variables is still modelled. The resulting capacities are often sparse in one or another representation and can be interpreted in terms of simplifying assumptions or limitations on the interactions that take place. We present the key definitions and properties, as well as implications for capacity-based aggregation.