A new model quantities the effect of soil structure, considered as the arrangement of particles in the soil, on soil water retention. The model partitions the pore space into texture-related and structure-related components, the textural component being what can be deduced to exist if the arrangement of the particles were random, and the structural component being the remainder. An existing model, based on particle-size distributions, represents the textural component, and a new model, based on aggregate-size distributions, represents the structural component. This new model makes use of generalized properties that vary little from one medium to another, thereby eliminating any need for empirically tilted parameters. It postulates a particular character of the structural pore space that in same ways resembles texture-related pore space, but with pore shape related to the breadth of the aggregate-size distribution. To predict a soil water retention curve, this model requires the soil's porosity and particle- and aggregate-size distributions. Tested with measurements for 17 samples from two sources, it fits the data much better than does a model based on texture alone. Goodness of fit indicated by correlation coefficients ranged from 0.908 to 0.998 for the new model, compared with a range of 0.686 in 0.955 for the texture-based model.