Subduction of oceanic lithosphere is widely considered as the primary cause of mantle heterogeneity as reflected by the chemical variation of ocean island basalts. However, the fate of deeply subducted oceanic lithosphere (i.e., remaining intact, getting segmented or completely destroyed and intermixed with ambient mantle) is still largely unknown. Based on trace element and radiogenic isotope systematics, it has been suggested that the geochemical difference of shield and post-erosional stage lavas of the volcanic Madeira archipelago reflects recycling of different portions of subducted lithosphere (e.g., Geldmacher and Hoernle, 2000, EPSL 183; Gurenko et al. 2013, Lithos 170-171). Accordingly, the geochemical composition of the shield stage magmas reflects altered upper oceanic crust, while the isotopically less enriched post-erosional magmas preferentially stem from less-modified, lower crustal/lithospheric mantle portions of the recycled slab. New, high precision potassium (δ41K of -0.75‰ to -0.50‰) and oxygen (δ18O of 4.90‰ to 5.21‰ in olivine phenocrysts) isotope data from Madeira lavas support this model. Subduction dehydration can cause large K isotope fractionation in the upper, seawater-altered parts of oceanic crust resulting in lighter K isotopes (lower δ41K) in ocean island basalts containing such material. The measured δ41K represents the lowest range among oceanic basalts published so far and overlaps with obducted eclogite, indicating the involvement of dehydrated oceanic crust. The correlation of light K with light O and enriched radiogenic isotope ratios supports the model that Madeira’s magma source contains different portions of subducted oceanic crust, which were preserved over millions of years without significant intermixing.