Two methods are compared for calculating the field-size dependence of the phantom scatter component of dose for x-ray beams. One model sums three Gaussian distributions; the other model is a two-parameter function. With a measurement of the beam quality as input to determine parameters, both models accurately reproduce the relative phantom scatter. However, there are important differences between the models. For all beam energies, the two-parameter model characterizes the absolute phantom scatter as a function of depth and field size, while, also for all beam energies, the six-parameter Gaussian model characterizes the relative phantom scatter at a single depth of 10 cm. For small field sizes, the phantom scatter calculated from the two-parameter model agrees with Monte Carlo calculations better than the Gaussian model. In the Gaussian model, the parameters can be obtained for beam energies between 60Co and 25 MV by linear interpolation based on the measured beam quality. In the two-parameter model, and for energies above 4 MV, the parameters can be obtained using linear functions of the dose-weighted average linear attenuation coefficient, which is related to beam quality.