The water-undersaturated melting relationships of a mafic, peralkaline, potassic madupite (with about 3% HzO as shown by chemical analysis) from the Leucite Hills, Wyoming, have been studied at pressures up to 30 kb. At low pressures (< 5 kb) leucite is the dominant liquidus phase, but it is replaced at higher pressures by clinopyroxene plus olivine (< 5-7 kb), clinopyroxene (7 12.5 kb), clinopyroxene plus minor spinel (12.5-17.5 kb), and clinopyroxene alone (17.5-> 30 kb). At all pressures there is a reaction relationship with falling temperature between melt, olivine and probably clinopyroxene to yield phlogopite. Apatite is stable within the melting interval to pressures above 25 kb. Electron microprobe analyses demonstrate that the clinopyroxene is diopsidic, with low aluminium and titanium contents. Pressure has relatively little effect on the composition of the pyroxene.. Phlogopite is also aluminium-poor and has only a moderate titanium content. The experimental results indicate that madupite is not the partial melting product of hydrous lherzolite or garnet lherzolite in the upper mantle and it seems improbable that it is derived by melting of mantle peridotite with a mixed HzO-CO 2 volatile component. Madupite could, however, be the partial melting product of mica-pyroxenite or mica-olivine-pyroxenite in the upper mantle. It is pointed out that the chemistry of some potassium-rich volcanics may have been affected by volatile transfer and other such processes during eruption and that experimental studies of material affected in this way have little bearing upon the genesis of potassic magmas. Finally, the experimental results enable constraints to be placed upon the P--T conditions of the formation of richteritebearing mica nodules found in kimberlites and asso-