Based on phase geochemistry and Re-Os isotopic ratios, an exotic (in an oceanic setting) K-rich silicate melt, named kimberlite-type, has been claimed to be the metasomatizing agent interacting with subcontinental lithospheric mantle fragments beneath the Cape Verde Archipelago. On the basis of textural features and major-and trace-element chemistry, we constrain key geochemical indicators able to discriminate percolation at depth of this exotic melt from infi ltration of the host magma in Cape Verde mantle xenoliths. Cape Verde type A lherzolites and harzburgites show evidence of dissolution of the primary phases (mainly pyroxenes) and the presence of large patches of secondary mineral (and glass) assemblages, and they do not show textural evidence of host basalt infi ltration. Cape Verde type A mantle xenoliths frequently contain almost pure K-feldspar (An3.8-8.8, Ab6-24, Or72-89) in the secondary mineral assemblage. They have an anomalously high K content (up to 0.49 wt%), and K/Na ratios generally >1, with respect to Cape Verde peridotites clearly affected by host basalt infi ltration (type B samples). The dichotomy between Na and K observed in the two textural types suggests that the Na-alkaline host basalt (K/Na <1), which infi ltrated at low pressure, was able to modify the whole-rock Na content of the xenoliths (type B samples). In turn, a completely different K-rich alkaline melt, which interacted at depth with the peridotite, imposed its alkali ratio (K/Na >1) on the bulk composition and formed the type A xenoliths. The kimberlitetype metasomatic agent, which reacted with the Cape Verde peridotite assemblage (mainly orthopyroxenes) in those regions where the mantle xenoliths are entrapped in the host basalt (P = 17 kbar; T = 1092°C), reasonably tends toward SiO 2-saturated, K-rich basic magmas (lamproite-type?) with K-feldspars as the "liquidus" phase. Isotopic data on separate clinopyroxenes do not contribute to discrimination between metasomatism and infi ltration processes but certainly concur to reinforce the hypothesis that a fragment of subcontinental lithospheric mantle domain was preserved during the opening of the Atlantic Ocean, forming K-rich undersaturated silicate melts that percolated through the peridotite matrix. Whole-rock major-and trace-element and isotopic geochemistry alone would not contribute to the interpretation of the processes occurring in the mantle xenolith. The most reliable tool would be an in situ mineral (and glass) study, which would be valid for Cape Verde mantle xenoliths and others. Small-melting-degree undersaturated silicate melts percolating at depth are olivine-saturated and may form, by reaction and dissolution of pyroxene, type A olivine without substantially modifying the original Fe/Mg peridotite ratio. By contrast, under low-pressure (<1.5 GPa), high-temperature regimes, olivine silicate melts infi ltrating the mantle xenoliths form type B olivine, in which Fe/Mg ratios will be controlled by fractionation. Mantle diopsides interact (at depth) with undersaturated silicate melts, rearranging the most fusible elements into a new diopside composition: type A clinopyroxene. By contrast, diopsides that interact with melts at progressively lower pressure react and are locally rearranged in a new chemical structure that is able to accommodate the high diffusive elements (i.e., Fe and Ti): type B aegirineaugites. Fe3+ in spinel is a key element in the investigation of the processes acting on Cape Verde mantle xenoliths. As a metasomatic product, secondary chromian spinel tends toward a Fe 3+-buffered composition, mainly depending on pressure and chemistry of the magma. A decompression system is able to change the percolation regime from porous flow to open conduit. At this stage, the chromian spinel would be the low-pressure phase able to accommodate larger amounts of Fe3+. Type A glasses have exceptionally high K2O content, and, when associated with K-feldspars, they are buffered at ~9 K2O wt%, in a silica range of 55.7-66.8 wt%. By contrast, type B glasses follow a hypothetical major-element trend toward the host basanites. In conclusion, the compositional features (in particular major elements) of minerals and glasses in relation to their chemical behavior in mantle systems are the most effi cient tools to distinguish metasomatism-related (type A) from infi ltration-related (type B) samples and consequently to place the mantle xenoliths in a correct genetic framework.