Petrographic and microprobe investigations of calc-alkaline (CA) rocks from the High Cascade Range (i.e., Mt. St. Helens, Mt. Jefferson, Crater Lake and Mt. Shasta) of western North America show that crystal clots represent primary igneous phase assemblages and are not products of amphibole reactions with melt. For each eruptive complex, crystal clots display diverse modal proportions even within a single eruptive unit. Nevertheless, in all cases the crystal-clot minerals are also represented in the rock as phenocrysts or microphenocrysts. Basalts contain clots of ol+plag+mgt, ol+mgt, cpx+ plag+mgt, cpx+mgt and plag+mgt; andesites, clots of cpx+mgt, opx+mgt, cpx+opx+plag+mgt, cpx+plag+mgt, opx+plag+mgt and plag±mgt; and dacites, clots of opx+mgt, cpx+opx+plag+ mgt, opx+plag+mgt, amph+plag+mgt±ilm, amph+mgt±ilm and plag±mgt. The bulk compositions of most of these clot assemblages could not have been derived from amphibole percursors. Although some amphiboles in dacitic rocks display a breakdown reaction of amph=plag+cpx+opx +mag, these mineral clusters, unlike those of clots, typically have a relict amphibole crystal outline and a fine-grained metamorphic texture. Plagioclase grains in the mineral clusters lack oscillatory zoning which is typical of crystal clot plagioclase grains. The euhedral to subhedral shapes of most clot minerals and the oscillatory zoning present in most clot plagioclase grains are not likely to have formed from the breakdown of amphibole. Crystal clots are also observed in Hawaiian and ocean floor basalts, although amphibole fractionation has not been proposed for those lavas. Magnetite fractionation may be the controlling process limiting iron enrichment in CA magmas rather than amphibole fractionation. Textural evidence indicates that magnetite is an early-forming phase in CA magmas. V, which is concentrated in magnetite, shows a strong decrease with increasing silica in many CA rocks, supporting a magnetite fractionation model.