Résumé : RF spoiling is a well established method to produce T1-weighted images with short repetition time gradient-echo sequences, by eliminating coherent transverse magnetization with appropriate RF phase modulation. This paper presents two novel approaches to describe signal formation in such sequences. Both methods rely on the formulation of RF spoiling as a linear increase of the precession angle between RF pulses, which is an alternative to the commonly used quadratic pulse phase scheme. The first technique demonstrates that a steady state signal can be obtained by integrating over all precession angles within the voxel, in spite of the lack of a genuine steady-state for separate isochromats. This clear mathematical framework allows a straightforward incorporation of offresonance effects and detector phase settings. Moreover, it naturally introduces the need for a large net gradient area per repetition interval. In a second step a modified partition method including RF spoiling is developed to obtain explicit expressions for all signal components. This provides a physical interpretation of the deviations from ideal spoiling behavior in FLASH and echo-shifted sequences. The results of the partition method in the small flip angle regime are compared with numerical simulations based on a Fourier decomposition of magnetization states. Measurements performed with in vitro solutions were in good agreement with numerical simulations at short relaxation times (T1/TR = 32 and T2/TR = 4), larger deviations occurred at long relaxation times (T1/TR =114 and T2/TR = 82).