The H₂O content of 35 glasses from Southwest Pacific back-arc basins (Lau, North Fiji, Woodlark and Manus) have been determined by infrared spectroscopy. On a plot of K₂O vs. H₂O the glass data define two distinct trends characterized by different slopes. Trend I, with a slope (K₂O/H₂O) of 0.25, can be explained by addition of a subduction-related component with K₂O/H₂O = 0.25 to a depleted mid-ocean ridge basalt mantle source (N- or D-MORB-like). Trend II, which coincides with the N- to E-MORB compositional spectrum, can be produced by addition of a non-subduction component, possibly an alkaline magma with K₂O/H₂O∼ 1.5, to the same depleted mantle source. The K₂O/TiO₂ and K/Nb values of E-MORB and back-arc basin basalts (BABB) of Trend II suggest that the enriched component involved in their genesis is not derived from a typical ocean island basalt (OIB, e.g. Hawaiian) mantle source. Our data show that the entire spectrum of BABB compositions can be explained by different degrees of mixing of a mantle source of either D-, N- or E-MORB composition with the subduction-related component, characterized by a K₂O/H₂O value of 0.25.

Different BABB types correlate with tectonic setting. Samples from the Trend II are associated with relatively stable spreading ridges, whereas those affected by the subduction-related component are always associated with more complex tectonic settings, or come from young or incipient back-arc basins. Pronounced E-MORB affinities of mantle sources are demonstrated only for samples from the Lau, North Fiji and Scotia Sea basins.

The most H₂O enriched BABB of Trend I partly overlap in terms of H₂O and K₂O content and H₂O/TiO₂ and K₂O/TiO₂ values with island arc tholeiites. This suggests involvement of similar subduction-related components in the genesis of these two magma types. Because a larger database is now available, the K₂O/H₂O vs. TiO₂ tectonic discriminan diagram of Muenow et al. [2] appears to be less useful than when originally proposed.

The very low K₂O/H₂O value ( < 0.05) of the H₂O-bearing phase involved in boninite genesis implies that it maybe a fluid derived from the subducted slab. The significantly higher K₂O/H₂O value (0.25) of the subduction-related component involved in petrogenesis of BABB and some arc tholeiites indicates that it was a melt, rather than a fluid. This K₂O/H₂>O value (0.25) is also of some interest, as the same value occurs in depleted MORB.