The isotopic compositions of ocean island basalts are usually taken to reflect the compositions of their mantle sources. We report Pb-Sr-Nd-O isotopic data for 52 samples, and chemical data for 70 samples from the major early Miocene subaerial shield construction phase of Gran Canaria, an ocean island located on the passive margin of northwest Africa. Small systematic differences in isotopic composition exist between the five vertical sequences of basalts sampled, with tight collinear Pb-Sr-Nd-O isotopic correlations in the lower parts of three sections, and very restricted isotopic compositions Pb-206/Pb-204-Pb-207/Pb-204 correlations are observed, with increases in Delta 7/4 Pb and Sr-87/Sr-86, and elsewhere. Negative decrease in epsilon(Nd), accompanied by increasing delta(18)O(cpx) values. These require that isotopic compositions of the basalts were primarily controlled by mixing between high-Pb-206/Pb-204 mantle and a crustal component. This crustal component must have been introduced into the magmas within the ocean island crust, since relatively large variations in delta(18)O(cpx) values (+ 5.2 to + 6.8 parts per thousand) are observed over a narrow range in Sr-87/Sr-86(0.70320 to 0.70390). The isotopic data of the lower parts of the sections can be precisely modelled by up to 8% bulk assimilation of NW African passive margin sediments by the least contaminated lavas: this is sufficient for 50% of the Pb in the most contaminated basalt to be sediment-derived, and to change substantially Ce/Pb, Ba/Nb and La/Nb ratios. No correlations exist between isotopic composition and fractionation indices, even in lavas that are closely spatially related, suggesting that contamination processes were not simple AFC-type processes. It follows that the absence of isotope composition-fractionation index relationships cannot be used as evidence that crustal contamination did not take place. It is very difficult to identify unequivocally an uncontaminated magma composition, since the least sediment-contaminated lavas, and most samples with Pb-206/Pb-204 > 19.55, have delta(18)O(cpx) values significantly below MORE and lunar values, and below mantle delta(18)O values determined by laser fluorination. Uncontaminated magmas could be derived from a mantle source with high Pb-206/Pb-204(>19.8), high Ce/Pb and negative Delta 7/4 Pb that resembles the compositions of the late Miocene undersaturated volcanics of Gran Canaria. If so, this source must have low delta(18)O values (less than or equal to + 5 parts per thousand), requiring an origin by recycling of hydrothermally-altered oceanic crust. Alternatively, the uncontaminated magmas could be derived from a mantle source on the Northern Hemisphere Reference Line with Pb-206/Pb-204 approximate to 19.6, which would imply that Pb-206/Pb-204 and low delta(18)O values, which are mostly quite evolved, acquired their characteristics by magmas with higher assimilation of hydrothermally-altered igneous crust with HIMU characteristics. The upper parts of the sampled stratigraphic sequences tend to be more evolved, may have elevated Ba and K abundances, and have very homogeneous more radiogenic Sr and Nd. These features are most simply explained by assimilation of altered MORE-type crust, or granitic bodies therein. If so, this implies that mantle-derived melts may have assimilated crust at three different levels in Gran Canaria (sediment, altered MORB-crust and altered HIMU-crust). In view of this, we would urge great caution in assuming that ocean island basalts reflect the isotope and trace element compositions of their mantle source, especially in areas of anomalously thick oceanic crust (e.g., Kerguelen) or close to continental margins.
