A fundamental question in the formation of orogenic andesites is whether their high melt SiO₂ reflects the recycling of silicic melts from the subducted slab or the processing of basaltic mantle melts in the overlying crust. The latter model is widely favoured, because most arc magmas lack the ‘garnet’ signature of partial slab melts. Here we present new trace element data from Holocene high-Mg# >64–72 calc-alkaline basalts to andesites (50–62 wt% SiO₂) from the central Mexican Volcanic Belt that crystallize high-Ni olivines with the high ³He/⁴He=7–8 of the upper mantle. These magmas have been proposed to be partial melts from ‘reaction pyroxenites’, which formed by hybridization of mantle peridotite (c. 82–85%) and heavy rare earth element-depleted silicic slab melt (>15–18%). Forward and inverse models suggest that the absence of a garnet signature in these melts reflects the efficient buffering of the heavy rare earth elements (Ho to Lu) in the subarc mantle. In contrast, all elements more incompatible than Ho – excepting TiO₂ – are more or less strongly controlled by the silicic slab flux that also directly contributes to the silicic arc magma formation. Our study emphasizes the strong link between slab recycling and the genesis of orogenic andesites.