Fig. 2. Comparison of is- land arc tonalite composi- tions from Fiji, Guadalca- nal, western North Ameri- ca, and Newfoundland A

16) with experimental re- 20-sults. There is excellent I kbar+ water agreement between closed n = 12 system experimental melt compositions and the nat- 6 ' 16 ural data, little agreement-between the 1-kbar water- 1 and saturated results and the 12- no wate natural data, and virtually n no overlap between the 3-kbar water-saturated ex-perimental melts and the 60 natural tonalites. All com-positions normalized to 100%, volatile-free; n = number of experiments. crustal pressures in a closed system. A likely place for such melting to occur is at or near the contact between hydrated arc crust and a hot basaltic intrusion. Local zones of melt-ing have been observed around basaltic in-trusions and elsewhere in several ancient island arc basement complexes (6, 14). Sig-nificant volumes of silicic magma can be generated by basalt-induced melting even of relatively mafic island arc crust (15), poten-tially enough to account for the observed abundance of island arc tonalite. Our experi-ments predict that dehydration melting will leave an anhydrous "granulitic " restite; rocks having the proper restite mineralogy have been described from at least one con-tact aureole (6). At water pressures as low as 1 kbar, there is significant divergence of experimental melt compositions and the nat-ural data set, and by 3-kbar water pressure, the compositions are distinctly different. In-dications from this and other studies are that the compositional gap between natural and experimentally produced melts increases as water pressure increases. Thus water-satu-rated melting of arc crust at pressures great-er than 1 kbar is an unlikely mechanism for the formation of island arc tonalite. 8 3 kbar + water n = 21 sradded &apos