Recently, Garrison and Davidson (2003) questioned the possibility that the adakites of the Northern volcanic zone of the Andes were generated by slab melting. The conclusions of the authors, involving a lower-crustal source for adakites, are in clear contradiction with ideas already proposed for the Northern volcanic zone and have implications not only for the understanding of the petrogenesis of this particular province but also more generally for adakite genesis. Based on a databank significantly wider than the one presented by Garrison and Davidson (2003), we observe different geochemical patterns, which lead to different conclusions.
The volcanic arc of Ecuador consists of three volcanic chains, which are clearly identified on the basis of geological, geochemical, and isotopic data. The western Volcanic Front and the Main Arc form the major chains, whereas backarc volcanism builds a smaller chain in the east. As a whole, volcanics consist of medium-K andesites and dacites, and subordinate rhyolites. Our databank consists of major and trace elements analyzed in ~1650 samples from 40 volcanoes, between latitudes 2.5°N and 2°S. Adakitic character is found in many volcanoes from the Main Arc and in almost all the edifices from the Front Arc (Bourdon et al., 2003). Moreover, Samaniego et al. (2002) showed that magmas in a single edifice changed in geochemical characteristics through time (from classical calc-alkaline to adakitic; e.g., Cayambe), reflecting the start of subduction of the Carnegie Ridge.
The samples used by Garrison and Davidson (2003) are almost exclusively from the Main Arc, and their study does not take into account these temporal constraints. Thus, the data set used by Garrison and Davidson is not representative of the Northern volcanic zone as a whole.
Garrison and Davidson (2003) assert that none of the Northern volcanic zone samples shows high-Mg andesite, yet high-Mg andesites have been found in several volcanoes of the Front Arc (e.g., Pichincha; Bourdon et al., 2003). If it is accepted that due to their high-MgO, Ni, and Cr contents, such rocks cannot be direct slab melts, geochemical and experimental work has shown that they can represent slab melts re-equilibrated with the mantle (Rapp et al., 1999). Consequently, we consider that the Mg, Ni, and Cr enrichments observed in some of the Northern volcanic zone lavas demonstrate adakitic melt–mantle peridotite interactions. Such exchanges may occur only if adakite source is located under a mantle slice, thus in the subducting slab (Martin and Moyen, 2002).
Garrison and Davidson (2003) use Sr/Y and silica contents to refute slab melting. The authors note that the Northern volcanic zone rocks overall have low Sr/Y ratios and stress that the highest values are from Sangay, a volcano of the Main Arc, precisely located above a dipping slab and probably not related to slab melting (Monzier et al.  noted that high Sr/Y values at Sangay result more from strong Sr enrichment than marked Y depletion). In addition, their model implies predicted Sr/Y values higher than 100 for adakites in Ecuador. This is not in agreement with the original definition: adakites have Sr/Y values higher than 40, whatever the geodynamic situation (Defant and Drummond, 1990).
Moreover, the SiO2 range of “putative slab melts” is assumed to represent the silica content of primary magmas produced in front of the Carnegie Ridge. Such an assumption should be valid only if all magmas represent true primary melts. However, fractional crystallization is an efficient process able to strongly modify silica content of magmas (including in the Northern volcanic zone). Consequently, silica definitely appears to be an inappropriate geochemical feature to distinguish slab melts.
Garrison and Davidson (2003) also argue that the lack of unequivocal geochemical variation along the arc excludes slab melting. However, data recently presented (Monzier et al., 2003) show systematic geochemical variation along the arc, all showing a negative or positive peak between 0.5°N and 1°S. Among those, Y and La/Yb display clear minimums and maximums, respectively, precisely where the Carnegie Ridge is subducting (Fig. 1). Such behavior reflects the intervention of slab melts in the petrogenesis of the magmas, directly related to the subduction of the Carnegie Ridge.
We agree with the Garrison and Davidson (2003) conclusion that the magma geochemical signature characterizes the source and not any specific geodynamic environment. The presence of adakites only indicates hydrous basalt melting, with a garnet ± amphibole residue. In Ecuador, the close spatio-temporal association of adakites, high-Mg andesites, adakite-like lavas, and high-Nb basalts (from front arc to backarc; Bourdon et al., 2003) is strong evidence of adakitic melt–mantle peridotite interaction, thus also pointing to slab melting.
The paper by Garrison and Davidson (2003) emphasizes the risk of making assumptions about regional geochemical variations based on an incomplete data set. Although discerning between slab and lower-crustal melting is challenging, we believe that Northern volcanic zone lava geochemical variations (across and along the arc) are best accounted for by slab melting beneath Ecuador. To discuss the petrogenesis of such a complex volcanic arc and try to approach the origin of its magmas, one needs to consider more than three geochemical characteristics (SiO2 being extremely weak). Only an exhaustive geochemical data set is able to show definitively the existence, or lack, of spatial and/or temporal variations.