The topic of the whistler precursor, a discrete VLF emission which is occasionally observed to preceed a whistler, is briefly reviewed. An analysis of some observations of precursors is presented, and two new mechanisms for the generation of precursors are described. It is shown that a mechanism in which two frequency components of a whistler, having the same group delay, interact to trigger a precursor at the difference of the frequencies does not predict results which agree well with observations. A mechanism in which part of the energy from the causative lightning stroke travels in an unducted mode, undergoing one magnetospheric reflection and refraction from the inner edge of the plasmapause, arriving in the equatorial plane with its wave vector parallel to the geomagnetic field, predicts results which are in good agreement with experimental observations. This method is compared with the model of Dowden (1972).