The extent of migration of residual magma during final solidification of a cumulate rock is difficult to determine, although various models and hypotheses have been proposed. In this article it is argued that the immiscible sulphide liquid accumulated syngenetically at the level of the Merensky Reef in the Bushveld Complex, and that the sulphide can be used as a tracer to study the migration of residual magma. The sulphide liquid would have had a very high density and low viscosity relative to silicate magma, and so should readily have permeated through its footwall if the latter was still a crystal mush with significant interstitial melt. This study of the chemical composition of the footwall rocks below the Merensky Reef shows that the sulphide liquid was only able to penetrate downward for distances of 1-2 m. It is suggested that if a high-density, low-viscosity sulphide liquid permeated only 1-2 m into its footwall, silicate liquid would have been even more restricted in its migration. The implication is that the footwall to the Merensky Reef had a low permeability by the time that the sulphide liquid in the Merensky Reef formed. Consequently, the concept of widespread, pervasive metasomatism by residual silicate magma is questioned, at least for the Bushveld Complex. In other, smaller intrusions in which crystal accumulation is more rapid, yielding a higher interstitial component, this argument may not be applicable. This study reports on one, thin sequence of rocks, and other studies at other levels in the intrusion would be desirable. The limitation is that an identifiable tracer (in this case a sulphide liquid) is required, which unequivocally monitors the process.