We present a method that can be used to estimate the amount of recycled material present in the source region of mid-ocean ridge basalts by combining three key constraints: (1) the melting behaviour of the lithologies identified to be present in a mantle source, (2) the overall volume of melt production, and (3) the proportion of melt production attributable to melting of each lithology. These constraints are unified in a three-lithology melting model containing lherzolite, pyroxenite and harzburgite, representative products of mantle differentiation, to quantify their abundance in igneous source regions.As a case study we apply this method to Iceland, a location with sufficient geochemical and geophysical data to meet the required observational constraints. We find that to generate the 20 km of igneous crustal thickness at Iceland's coasts, with 30±10%30±10% of the crust produced from melting a pyroxenitic lithology, requires an excess mantle potential temperature (ΔTp) of ⩾130 °C (Tp⩾1460°C) and a source consisting of at least 5% recycled basalt. Therefore, the mantle beneath Iceland requires a significant excess temperature to match geophysical and geochemical observations: lithological variation alone cannot account for the high crustal thickness. Determining a unique source solution is only possible if mantle potential temperature is known precisely and independently, otherwise a family of possible lithology mixtures is obtained across the range of viable ΔTp. For Iceland this uncertainty in ΔTp means that the mantle could be >20% harzburgitic if ΔTp>150°C (Tp>1480°C).The consequences of lithological heterogeneity for plume dynamics in various geological contexts are also explored through thermodynamic modelling of the densities of lherzolite, basalt, and harzburgite mixtures in the mantle. All lithology solutions for Iceland are buoyant in the shallow mantle at the ΔTp for which they are valid, however only lithology mixtures incorporating a significant harzburgite component are able to reproduce recent estimates of the Iceland plume's volume flux. Using the literature estimates of the amount of recycled basalt in the sources of Hawaiian and Siberian volcanism, we found that they are negatively buoyant in the upper mantle, even at the extremes of their expected ΔTp. One solution to this problem is that low density refractory harzburgite is a more ubiquitous component in mantle plumes than previously acknowledged.