Trace elements are widely used to interpret the origin of mantle-derived magmas, yet we lack detailed understanding of how trace elements behave during melting of mantle source components. Here, we present new data on trace element distribution and partitioning between phases from high pressure (3.0 to 5.0 GPa), high-temperature (1230 to 1550 °C) melting experiments on starting compositions that represent altered oceanic crust and metasedimentary protoliths. These compositions are expected to be recycled into the mantle via subduction or delamination to form heterogeneous mantle domains that are implicated in the genesis of intraplate and/or ocean floor magmas. In most of the experiments, the investigated trace elements behave incompatibly, expect for HREE and Y, which are compatible in garnet, and V, Cr and Zn, which partition into both garnet and clinopyroxene. Relative to Nd, P is more compatible in garnet than clinopyroxene, leading to fractionation of P/Nd with melting in some cases. Melt compositions in some experiments with low melt fractions feature distinctive negative anomalies for Nb, and for Sr, Ba and Eu, due to retention of these elements in minor/accessory rutile and feldspar, respectively. We also show that highly incompatible trace element (e.g., Cs, Th, U, LREE) concentrations in melts are strongly controlled by melt fraction, whereas moderately incompatible (M-HREE, Zr) to compatible (Cr, V) element concentrations are controlled by temperature and/or phase composition. Pressure has relatively little influence on trace element behaviour at the investigated conditions. Based on our results, we suggest that partial melting of eclogitic components of mantle domains may ultimately produce magmas with trace element compositions that are unlike peridotite-sourced magmas. Therefore, the trace element systematics of mantle-derived magmas should not only be interpreted in terms of mantle source compositions, but also with consideration to source petrology (e.g., mineral compositions and accessory phase stability) and melting conditions (e.g., melt fraction, pressure, temperature). © 2017 Elsevier B.V.