Muscle myosin filament backbones are known to be aggregates of long coiled-coil cu-helical myosin rods, but the packing arrangement is not understood in detail. Here we present new data on fish muscle myosin filaments from low-angle X-ray diffraction and from freeze-fracture, deep-etch electron microscopy which put constraints on the kind of models that might explain all of the observations. In particular, it is known in the case of vertebrate striated muscle thick filaments that the myosin head array in resting muscle is not perfectly helical but contains periodic perturbations. We show by analysis of low-angle X-ray diffraction patterns from resting bony fish muscle that any radial, azimuthal, and axial perturbations of the myosin head origins on the filament surface (due to perturbed myosin rod packing) must all be rather small and that the main perturbations are in the myosin head configurations (i.e., tilts, slews, rotations) on those origins, We provide evidence that the likely arrangement of titin molecules on the myosin filament is with them aligned parallel to the filament long axis, rather than following helical tracks, We also show from freeze-fracture studies of fish muscle that the myosin filament backbone (including titin and other extra proteins) has a radius of about 65-75 Angstrom and appears to contain a small (approximate to 15-20 Angstrom radius) hollow core. Together with previously published evidence showing that the myosin rods are nearly parallel to the thick filament long axis, these results are consistent with the curved crystalline layer model of Squire (J, M. Squire, 1973, J. Mol. Biol. 77, 291-323), and they suggest a general structure for the C-zone part of the thick filament