Crystal clusters are common in both extrusive and plutonic rocks, but the mechanisms by which they form are not well-constrained. Following a consideration of the physics of nucleation, we outline the expected microstructural characteristics of clusters formed by heterogeneous nucleation and those formed by synneusis, together with the ways they might evolve during subsequent grain growth and textural equilibration. By combining analysis of the microstructures in experimental chromite-basalt charges with a detailed microstructural analysis of the UG2 chromitite of the Bushveld layered intrusion using EBSD, we argue that the UG2 chromitite formed by settling and accumulation of single grains and clusters comprising randomly oriented grains produced by the aggregation of previously isolated chromite crystals. Although there is no evidence of epitaxy, at least some of the lowermost chromite grains of the main UG2 chromitite may have nucleated heterogeneously on the silicate grains forming the floor, with subsequent accumulation and sintering of individual grains or clusters. The reduced thickness of chromitites on the steep and overhanging parts of the floor is thus due to the relative difficulty of sticking more grains to the existing layer in these orientations. The absence of any fining-upwards of grains in either the main UG2 chromitite or the associated stringer can be accounted for if both layers were formed by the settling and accumulation of clusters as well as single grains. Comparison with examples of clustered chromite grains in extrusive rocks suggests that aggregation by synneusis is a widespread magmatic process. The “chicken-wire” texture formed by clustered chromite grains commonly found in olivine-rich cumulates is argued to also be formed by gravitational settling, with the possible exception of clusters of chromite grains in relatively thin seams argued to be the result of metasomatism, which may instead have formed by impingement during in situ growth.