Here, we critically examine models for the development of layering in the ultramafic rocks of the Stillwater Complex of Montana, USA. A fine-grained, high-magnesium (high-Mg) dike in the Mountain View area of the complex is a plausible example of the type of magma that was parental to the layered cumulate rocks in the ultramafic series. Modeled assimilation and crystallization of a primary komatiitic magma, involving 20% contamination by Archean granodiorite and 1% iron formation in a batch process, produces a liquid and solid assemblage like that in the dike. Equilibrium crystallization during assimilation of the crust produces cumulus olivine (Fo87-85) and chromite (Cr#=64-55) consistent with the mineral chemistry of cumulus phases in the Peridotite Zone; each internally differentiated ultramafic layer could have formed by injection of a single highly contaminated batch of komatiite magma carrying about 10-20 % suspended crystals. Whereas previous models of fractional crystallization to form the limited range of mineral and bulk-rock compositions observed in the Ultramafic Series required that these rocks accumulated from tens to thousands of times greater thicknesses of melt, the present model requires only 4 times as much melt as cumulates and dramatically diminishes the apparent volume of missing magma. The modeled crystallization of the proposed parental magma also demonstrates that, relative to olivine, only a small amount of cumulus chromite crystallizes in cotectic volume ratios of around 100:1 to 100:2 olivine to chromite. The modal abundance of chromite in the olivine-bearing cumulates from the Mountain View area, varies continuously between 0.1 and 80 vol%, with an average ratio of 100:9.2 olivine to chromite, rarely showing proportions consistent with either chromite-only or chromite + olivine cotectic crystallization. This, in turn, suggests that if olivine and chromite did precipitate in cotectic proportions, the crystals have subsequently been sorted mechanically to variable extents, producing the observed continuous range of modal proportions of chromite in the cumulates, and olivine has been preferentially removed from the system along with considerable amounts of escaped melt. Furthermore, the behavior of crystallizing assemblages at the olivine-orthopyroxene and the chromite-orthopyroxene peritectics forbids the deposition of the commonly observed mineral assemblages olivine + orthopyroxene + chromite or orthopyroxene + chromite by a process of fractional crystallization, but does permit them to form by deposition of mechanically sorted crystals derived from magmas carrying large crystal loads at internal equilibrium. Extremely efficient sorting of crystal-rich magma during lateral transport of crystals may therefore be responsible for the nearly monomineralic chromite seams in the Peridotite Zone. These considerations lead us to question the purported cyclicity of the cumulate rocks in the Peridotite Zone. Markov Chain analysis of the layering in logged drill core at various locations in the complex show considerable departure from the suggested regular cycles of cumulate layers and provides no support for the classic model that units formed by closed-system equilibrium crystallization from mixed magmas following a succession of discrete magma recharge events. We suggest that the degree of contamination in the supply of crystal-laden contaminated magma transported to the site of deposition may have varied randomly, leading to the formation of layered rocks that do not conform to the previously accepted cyclic unit model of settled or in situ liquidus minerals. The layered rocks observed in the Peridotite Zone may have formed at the base of a magma chamber or they may not have, but their compositions and mineral modes do not convey any information about such a magma body if it was present at the time of their formation, because each layered unit appears to have formed from a completely separate volume of komatiitic magma.