The petrology and geochemistry of Icelandic basalts have been studied for more than a century. The results reveal that the Holocene basalts belong to three magma series: two sub-alkaline series (tholeiitic and transitional alkaline) and an alkali one. The alkali and the transitional basalts, which occupy the off-rift volcanic zones, are enriched in incompatible trace elements compared to the tholeiites, and have more radiogenic Sr, Pb and He isotope compositions. Compared to the tholeiites, they are most likely formed by partial melting of a lithologically heterogeneous mantle with higher proportions of melts derived from recycled oceanic crust in the form of garnet pyroxenites compared to the tholeiites. The tholeiitic basalts characterise the mid-Atlantic rift zone that transects the island, and their most enriched compositions and highest primordial (least radiogenic) He isotope signature are observed close to the centre of the presumed mantle plume. High-MgO basalts are found scattered along the rift zone and probably represent partial melting of refractory mantle already depleted of initial water-rich melts. Higher mantle temperature in the centre of the Iceland mantle plume explains the combination of higher magma productivity and diluted signatures of garnet pyroxenites in basalts from Central Iceland. A crustal component, derived from altered basalts, is evident in evolved tholeiites and indeed in most basalts; however, distinguishing between contamination by the present hydrothermally altered crust, and melting of recycled oceanic crust, remains non-trivial. Constraints from radiogenic isotope ratios suggest the presence of three principal mantle components beneath Iceland: a depleted upper mantle source, enriched mantle plume, and recycled oceanic crust. The study of glass inclusions in primitive phenocrysts is still in its infancy but already shows results unattainable by other methods. Such studies reveal the existence of mantle melts with highly variable compositions, such as calcium-rich melts and a low- 18 O mantle component, probably recycled oceanic crust. Future high-resolution seismic studies may help to identify and reveal the relative proportions of different lithologies in the mantle.