The width of the transition region between oppositely magnetized regions in nonreturn-to-zero (NRZ) digital recording is calculated by minimizing the energies involved, taking into account to a first approximation the interaction with the read-head. The writing process is neglected and argued to be relatively unimportant. The width is found to be in agreement with all available experimental data on pulse width when the binary digits are long (low-recording packing densities). The average field in the magnetized regions at high-packing densities is found to be rather well approximated by a field of a point dipole, thus justifying the previous calculations (which assumed such a dipole field) whose results for the maximum packing density also agree with experiment. The signal amplitude at low-recording packing densities, calculated from the same potentials, fits the experiment rather well, especially if cubic, rather than uniaxial, anisotropy is assumed. It seems that a better fit might be obtained if a better approximation is used for the interaction with read-head.