Experiments were conducted to explore the behavior of Li, Rb, Nb, Sn, Cs, Ta, W during crustal melting and test the anatectic origin of rare metal-bearing peraluminous granites such as rare metal granites (RMGs). The experiments were performed under fluid-absent conditions at 800 and 850 °C, 400 MPa and moderately reducing fO 2 (ΔFMQ = − 0.5 to − 0.8). Starting materials were cores of several millimetres drilled from two natural rocks, a biotite-rich paragneiss (Pg) and a muscoviterich orthogneiss (Og) enriched in Li, Be, Sn, Cs, W. Both protoliths produced small melt fractions from 8 to 20% vol. Melt distributions were either homogeneously distributed at grain boundaries in the Pg or preferentially associated with muscovite reaction zones in the Og. In the Pg at 800 °C, melting is mainly fluid present, driven by interstitial water at grain boundaries. At 850 °C, biotite dehydration-melting produces peritectic orthopyroxene, hercynitic spinel, ilmenite and alkali feldspar in addition to melt. In the Og, muscovite dehydration-melting generates melt plus peritectic biotite, hercynitic spinel, ilmenite, Al silicates and alkali feldspar. Experimental glasses are nearly homogeneous, silica rich, peraluminous and leucogranitic and their major element compositions differ only little between the two protoliths. In contrast, the trace element concentrations vary as a consequence of chemical and textural heterogeneities in our starting materials. Compared with source rocks, the Og glasses are enriched in Rb, Nb, Ta, W and depleted in Li, Cs and the Pg are enriched in Li, Rb, Cs, W and depleted in Nb, Ta. Mass-balance calculations indicate that during muscovite dehydration-melting, Li, Cs and Rb partition into the melt; whereas Nb, Ta and W are preferentially incorporated in peritectic phases. Li and Cs also partition toward the melt during biotite dehydration-melting. The partitioning behavior of trace elements during crustal melting is a function of the melting reaction and partition coefficients between melt, residual and peritectic phases. Experimental glasses are similar to peraluminous muscovite granites but fail to reproduce RMG compositions. Alternatives to mica dehydration-melting such as fluid-present and residual source melting emphasize the difficulties with an origin of RMGs by purely anatectic processes. Crystallization differentiation might have to be combined with mica dehydration-melting to explain the distinctive geochemical features of RMGs.