Muscle myosin filament backbones are aggregates of long coiled-coil alpha-helical myosin rods, with the myosin heads arranged approximately helically on the filament surface, but the details of the rod packing are not known. Computed Fourier transforms of plausible molecular packing models for the vertebrate striated muscle myosin filament have been compared with observed high-angle X-ray diffraction patterns from plaice fin muscle. Models considered include those in which the coiled-coil rod parts of myosin are packed into various kinds of subfilaments or into a curved molecular crystalline layer. A general conclusion is that if the myosin rods are tilted by less than about 1 degree or more than about 3 degrees from the filament long axis, very poor agreement is obtained between the computed and observed high-angle diffraction patterns. Qualitative comparison of calculated Fourier transforms, taken together with electron micrograph information, shows that the curved molecular crystal model and a model with hexagonally close-packed 4-nm subfilaments appear to explain the whole set of observations more satisfactorily than the alternatives. It is argued on other grounds that of these two possibilities the curved molecular crystal model is the more plausible.