The thickness of the crystal mush on magma chamber floors can be constrained using the offset between the step-change in the median value of dihedral angles formed at the junctions between two grains of plagioclase and a grain of another phase (typically clinopyroxene, but also orthopyroxene and olivine) and the first appearance or disappearance of the liquidus phase associated with the step-change in median dihedral angle. We determined the mush thickness in the Rustenburg Layered Suite of the Bushveld Complex at clinopyroxene-in (in Lower Main Zone) and magnetite-in (in Upper Zone). We also examined an intermittent appearance of cumulus apatite in Upper Zone, using both the appearance and disappearance of cumulus apatite. In all cases, the mush thickness does not exceed 4 m. These values are consistent with field observations of a mechanically rigid mush at the bases of both magnetitite and chromitite layers overlying anorthosite. Mush thickness of the order of a few metres suggests that neither gravitationally-driven compaction nor compositional convection within the mush layer is likely to have been important processes during solidification: adcumulates in the Bushveld are most likely to have formed at the top of the mush during primary crystallisation. Similarly, it is unlikely either that migration of reactive liquids occurs through large stretches of stratigraphy, or that layering is formed by mechanisms other than primary accumulation.