Potassic alteration domains of Archean greenstone belt lode gold deposits, typified by the Hollinger–McIntyre and Buffalo Ankerite mines, Timmins, are characterized by systematic partitioning between different groups of incompatible elements such that Al, Ga, Th, U, Ti, and V are decoupled from K, Rb, Ba, Li, Cs, and Tl. The former group of elements has concentrations and interelement ratios in alteration domains that reflect host-rock control, implying approximately isochemical behaviour. In contrast, the lithophile elements K, Rb, and Ba are generally coenriched and are linearly correlated over three orders of magnitude in abundance, where K/Rb = 230–380 (r = 0.92–0.99) and K/Ba = 34–97 (r = 0.82–0.99), values close to average crustal ratios of 285 and 36, respectively. K/Rb and K/Ba ratios trend towards higher values with respect to increasing concentrations of Rb and Ba, possibly because of mixing between host rock and hydrothermal reservoirs of the lithophile elements. Cs and Tl contents average 1.4 and 1.1 ppm, respectively, such that K/Cs and K/Tl, as well as K/Rb and K/Ba, trend to higher than bulk crustal ratios, signifying a mafic to ultramafic source variably depleted in Cs, Tl, Rb, and Ba relative to K. Lithium is generally not coenriched with K but rather correlates with Mg content, reflecting the primary magmatic substitution of Li1+ for Mg2+ in the parent magmas. Al and Ga are highly correlated: Al/Ga = 2070–3790 and r = 0.85–0.98. Accordingly, Al and Ga were not fractionated during mineralization. These interelement trends, collectively, are present in deposits variously hosted by ultramafic, mafic, or felsic volcanic rocks and by sediments or granitoids.Magmatic processes involving crystal fractionation of biotite, K-feldspar, and plagioclase generate trends to systematically diminished K/Rb (≥50), K/Li, K/Cs, and K/Tl but enhanced K/Ba (≤8 × 103) and Rb/Sr ratios in most late-stage differentiates. Al/Ga ratios diminish systematically during late-stage magmatic evolution, and Th–U and Ti–V display mutual partitioning governed by crystallization of trace phases. Such "late-stage" trends are the rule in "magmatophile" deposits, including the Archean Cadillac molybdenite deposit, Phanerozoic Cu- and Mo-"porphyry" deposits, Sn–W greisens, and most pegmatites. Accordingly, magmatic processes of this type can be ruled out as the dominant source of volatiles for gold-forming systems. There is no complementarity between large-ion lithophile-element (LILE) (K, Rb, Ba, Cs, U, Th) depletion patterns in granulites and K, Rb, and Ba enrichment in Au deposits. The enhanced K/Rb, K/Cs, and Rb/Cs ratios but diminished K/Ba ratios, characteristic of many granulites, are not reflected in the gold deposits, where K/Rb and K/Ba ratios approximate average crustal values and Th and U behave isochemically. In many granulites, especially Archean examples, LILE depletion is a primary feature; accordingly, the granulitization model for Au–Ag vein deposits can also be ruled out.The compliance of K/Rb and K/Ba ratios in potassic alteration domains of Au deposits with values characteristic of main-trend igneous rocks or "average" crust implies that K, Rb, and Ba were partitioned into the hydrothermal ore-forming fluids in approximately the same ratios as in the source rocks, consistent with known fluid–mineral distribution coefficients. Dehydration reactions in dominantly mafic to ultramafic source rocks under amphibolite-facies conditions or equilibration of ore-forming fluids with source rocks under conditions of low water/rock ratio, rather than either purely magmatic or granulitization processes, may satisfy the requirement for proportional K, Rb, and Ba coenrichment, Li retention, and the absence of Al–Ga fractionation.