We have re-evaluated mantle potential temperature estimates for the North Atlantic Igneous Province (NAIP). Temperature estimates involving olivine addition to pillow-lava glasses are unreliable because host glasses formed along the liquid+olivine+plagioclase cotectic and not just the olivine liquidus. Additionally, magma chamber processes can generate picritic lavas containing only magnesian olivine, but picrites alone do not require high mantle temperatures. Furthermore, petrological models tend to overestimate TP in picrites containing appreciable accumulative olivine further confusing the issue. Selected aphyric lavas from West Greenland, which cannot have accumulated olivine, suggest maximum TP~1500°C. Petrological models for Icelandic glasses suggest a maximum TP~1450° C which is consistent with olivine-melt and olivine-spinel equilibration temperatures. However, melting of 'damp' peridotite beneath Iceland would reduce this estimate perhaps by 50°C. The NAIP mantle was lithologically and chemically heterogeneous and was made of a hybrid pyroxenite-peridotite lithology, the pyroxenite component being derived from recycling of subducted slabs. However, there is no necessity for the subducted slabs to have been recycled to the core-mantle boundary. Pyroxenite could have been derived from Caledonian-aged slabs that also hosted helium with high 3 He/ 4 He within the shallow mantle, which was inherited by Palaeocene or young melts. The pyroxenite component was more readily fusible than the peridotite component under the same P-T conditions, allowing variations in melt production rate throughout the province. Melting of lithologically variable mantle is consistent with observed radiogenic isotope variability in Icelandic basalts and related trace-element variations in throughout the NAIP. We propose that magmatism in the NAIP resulted from extensional tectonics above 'warm' mantle that had been internally heated beneath thick continental lithosphere prior to continental break up. Only in areas of extension did magmatism occur, thus explaining the apparently widespread initial phase of magmatic activity.