The Bushveld Complex is traditionally interpreted as having formed from numerous magma pulses that successively replenished the evolving chamber and deposited crystals on a temporary floor, building the layered sequence from the base upwards. There are, however, several historic and recent attempts to question this view, with the suggestion that some of the layers (e.g. UG2 chromitite layer) or even entire units (e.g. Merensky Unit) are late-stage sills intruded into the nearly solidified cumulate pile. These alternative models would require that the 'sill-like layers/units' reveal cross-cutting relationships with their host rocks, in particular, with their overlying cumulates. Compelling field evidence for such relationships has, however, never been presented. Contrary to this prediction, we show here that the immediately overlying rocks commonly cut down through the inferred sill-like layers/units. This eliminates the possibility that these layers/units represent sill-like intrusions into a pre-existing cumulate pile. In addition, sill emplacement is a random mechanism that should result in rock sequences with chaotic-looking compositional trends. This prediction is at odds with the systematic cumulate stratigraphy of the Bushveld Complex. We thus conclude that attempts to reinterpret some layers/units as late-stage sills are poorly grounded, and we reconfirm that sequential upward deposition of layers on the chamber floor is the most reasonable explanation for most features of the Bushveld Complex.