The late Pleistocene Lake Basalt of Medicine Lake volcano, California is comprised of variably porphyritic basalt and basaltic andesite flows and scoria. These eruptives are similar in composition and phenocryst abundance to the low-MgO, high-Al²O³ mafic magmas common in convergent margin settings. The petrogenesis of the magmas that produced the Lake Basalt has been inferred from field relations, melting experiments and subsequent major and trace element modeling. Their formation involved both hydrous differentiation and plagioclase accumulation and thus the Lake Basalt can be used to constrain the relative contributions of these processes to the production of high-Al²O³ arc basalt. Phenocryst-poor lavas of the Lake Basalt formed by hydrous differentiation; their compositions and observed phenocrysts were reproduced in 1 kbar, H²O-saturated melting experiments. Anorthite-rich plagioclase compositions of the lavas of the Lake Basalt necessitate crystallization from melts with between 4 and 6 wt% dissolved H²O. Phenocryst-rich lavas of the Lake Basalt, with 18 modal% phenocrysts and greater, formed by plagioclase accumulation in magmas similar to the phenocryst-poor lavas. This interpretation is supported by the depleted incompatible element abundances and enriched Sr/Zr ratio of the more porphyritic lavas relative to the phenocryst-poor lavas. We model the formation of the Lake Basalt as a two-stage process that combines a differentiation model and a plagioclase accumulation model. Stage one involved hydrous fractionation, granitic assimilation and mixing with undifferentiated parent magma. This process generated lavas with up to 19.2 wt% A1²O³ and 7 modal% phenocrysts. In stage two, plagioclase accumulated in these liquids and produced more aluminous and porphyritic lavas with up to 21.8 wt% A1²O³ and 33 modal% phenocrysts.