In order to represent hysteretic soil water retention curves accurately using as few measurements as possible, a new semiempirical model has been developed. It has two postulates related to physical characteristics of the medium, and two parameters, each with a definite physical interpretation, whose values are determined empirically for a given porous medium. One parameter represents the fraction of the pore space that is not subject to Haines jump hysteresis. Its associated postulate is that a single value of this parameter characterizes a given medium, or, equivalently, that the medium is self-similar with respect to the division between hysteretic and nonhysteretic pore space. The second parameter is the effective body-to-neck size ratio of the medium's largest pore. The second postulate specifies a particular relation between the size distributions of pore bodies and of pore necks. Tests of the model show that it provides high-quality optimized fits to measured water content vs. matric pressure wetting curves for a wide variety of media. The parameter values obtained through optimization correlate plausibly with such media properties as uniformity of particles, complexity of structure, and degree of compaction. A practical use of this model is to provide a complete simulated main wetting curve for a medium where only a main drying curve and two points on the wetting curve have been measured. With additional development, it may be possible to do without the two measured wetting points if parameters can be evaluated from other measurements or known properties of the medium.