The data of Joesten (1986) are re-interpreted. The petrography of the coronas is not consistent with magrnatic origin. Both microstructural types described by Joesten (1986), here re-named ‘columnar’ and ‘tabular’, formed by solid-state replacement of plagioclase and of adjacent olivine or ilmenite. Tabular microstructures are not annealed, but result from overgrowth or epitaxy of amphibole and pyroxene on pre-existing grains. Since the diffusion-controlled models of Joesten (1986) can account for major aspects of the coronas, it seems possible that a slightly modified, less simplified theory might explain them fully. Open-system behaviour must be admitted, with some constraints provided by symplectites. It may also be necessary to develop the theory in more than one dimension, and to allow for departures from local equilibrium at layer boundaries.

Coronas of symplectic pargasite [XMg = 68, XCa = 69, XSi = 68] + spinel [SP45] and orthopyroxene [EN70] separate cumulus plagioclase [AN64] and olivine [FO64] in troctolitic gabbro from Risör, Norway. Coronas with a primary microstructure characterized by low-angle grain boundaries have the mineral assemblage layer sequence,

PLAGIOCLASE : PARGASITE + SPINEL : ORTHOPYROXENE + SPINEL : ORTHOPYROXENE : OLIVINE, whereas those with an annealed microstructure have the layer sequence,

PLAGIOCLASE : PARGASITE + SPINEL : PARGASITE : ORTHOPYROXENE : OLIVINE.

Primary symplectite consists of fan-shaped grains of pargasite that open toward plagioclase and are ribbed by rods of spinel. On annealing to a single grain, low-angle grain boundaries disappear and spinel rods rotate into parallel positions. Primary orthopyroxene, initially radial to layer contacts, anneals to a single grain, rimming olivine. Irregular discontinuous layers and plumes of symplectic orthopyroxene+spinel disappear on annealing and a monomineralic layer of pargasite forms between orthopyroxene and the amphibole symplectite.

The identity of composition of intercumulus pargasite and orthopyroxene with the same phases in primary coronas, the occurrence of primary and annealed coronas rimming cumulus ilmenite and the occurrence of radial orthopyroxene separating grains of cumulus olivine and separating cumulus olivine and plagioclase argue for the origin of primary coronas by fractional crystallization of spinel-saturated troctolitic magma at a pressure greater than 5 kbar.

Irreversible thermodynamic calculations show that the mineral assemblage layer sequence of the primary coronas is diffusionally unstable and that this instability provides the driving force for their spontaneous transformation to the stable mineral assemblage layer sequence of the annealed coronas.