For a number of years it has been supposed that for multidomain materials, the demagnetization factor N can be determined from either hysteresis loop or initial susceptibility measurements. Unfortunately, these methods involve a subtle circular argument which invalidates the entire analysis, ruling out the use of either of the aforementioned measurements in the determination of N. On examining possible alternatives to these methods we find the only viable one to be ferromagnetic resonance spectroscopy. Data are presented from a single crystal of magnetite which demonstrate that a reasonable value for N can be derived although only for the saturated state.