Hydrothermal experiments were carried out at 2 kbar water pressure, 700 °–800 ° C, with the objective of determining the level of dissolved Zr required for precipitation of zircon from melts in the system SiO2-Al2O3-Na2O-K2O. The saturation level depends strongly upon molar (Na2O + K2O)/Al2O3 of the melts, with remarkably little sensitivity to temperature, SiO2 concentration, or melt Na2O/ K2O. For peraluminous melts and melts lying in the quartz-orthoclase-albite composition plane, less than 100 ppm Zr is required for zircon saturation. In peralkaline melts, however, zircon solubility shows pronounced, apparently linear, dependence upon (Na2O + K2O)/Al2O3, with the amount of dissolvable Zr ranging up to 3.9 wt.% at (Na2O + K2O)/Al2O3 = 2.0. Small amounts (1 wt.% each) of dissolved CaO and Fe2O3 cause a 25% relative reduction of zircon solubility in peralkaline melts. The main conclusion regarding zirconium/zircon behavior in nature is that any felsic, non-peralkaline magma is likely to contain zircon crystals, because the saturation level is so low for these compositions. Zircon fractionation, and its consequences to REE, Th, and Ta abundances must, therefore, be considered in modelling the evolution of these magmas. Partial melting in any region of the Earth's crust that contains more than ∼ 100 ppm Zr will produce granitic magmas whose Zr contents are buffered at constant low (< 100 ppm) values; unmelted zircon in the residual rock of such a melting event will impart to the residue a characteristic U- or V-shaped REE abundance pattern. In peralkaline, felsic magmas such as those that form pantellerites and comendites, extreme Zr (and REE, Ta) enrichment is possible because the feldspar fractionation that produces these magmas from non-peralkaline predecessors does not drive the melt toward saturation in zircon. Zircon solubility in felsic melts appears to be controlled by the formation of alkali-zirconosilicate complexes of simple (2:1) alkali oxide: ZrO2 stoichiometry.