Abstract: New field observations of certain harrisite occurrences are presented, including: interruption of layers; splitting of very thick layers into several smaller ones; existence of small isolated lenses and pods of harrisite and of upwardly extending tongues of harrisite; harrisite forming parts of the matrix of breccias; and isolated pods of harrisite along the western margin of the intrusion. These layers, lenses, tongues, and pods seem to have crystallized within the cumulus mush rather than at the mush magma boundary. It is proposed that the rock in these instances represents segregated, intercumulus melt which ‘ponded’ beneath relatively impermeable layers in the cumulus mush. Several ways in which supercooling may have arisen to cause skeletal olivine growth are considered and previous estimates are thought to need reduction by 10–20°C. It is suggested that segregation of upward-filtering melt in other layered intrusions might produce layers indistinguishable from ‘uniform’ cumulate.