It is generally agreed that the driving force (plastic strain energy) is much too small to allow "classical" nucleation during static and dynamic recrystallisation, and that rotation/growth of subgrains is an alternative. The latter explanation predicts that new grains should begin at low angles to old grains. We have used electron backscatter diffraction on an experimentally deformed quartz polycrystal that has deformed by dislocation creep and partially recrystallised. In a region shortened by about 30% new grains are at high angles (much greater than 15degrees) to adjacent parent grains. A histogram of misorientation versus number of boundaries shows a gap at 15-20degrees. In its simple form we expect the subgrain rotation model to predict a spectrum of misorientations but with most of them being low angle. Instead, the histogram suggests that new boundaries began life as high-angle structures, so current models for deformation-induced nucleation require refinement.