An amphibolite has been experimentally deformed under subsolidus and partially molten conditions to evaluate the influence of melt on the mechanical behavior of a natural mafic rock and to assess low-melt fraction segregation processes. Experiments were performed under fluid-absent conditions at 1.8 GPa, between temperatures of 650 ¿C and 1000 ¿C. These conditions are similar to those of thickened lower continental crust or intermediate depths in subducting oceanic lithosphere. At ≥850 ¿C, melt is granitoid in composition, and its viscosity is that of ''wet granite'', 103--105 Pa s. The results can be summarized as follows: (1) Under subsolidus conditions (tests at 650 ¿C and 750 ¿C the amphibolite is macroscopically ductile and deformation is homogeneously distributed throughout the sample. (2) At near solidus conditions (≥800 ¿C, ~0--5 vol % melt), fractures (~1--10 μm in width) displace hornblende and plagioclase grains, and melt, formed in situ, is found in some of these cracks. The formation of a ductile shear zone in one sample is attributed to the presence of very fine grained reaction products from combined dehydration/hydration reactions that involve plagioclase. The dehydration reaction products have apparently changed the deformation mechanism and even overrode the melt-embrittlement process, trapping melt in pockets of lower strain. With higher melt fractions (~10--15 vol %), broad, melt-bearing shear zones form and grains within these zones are brittlely deformed.

(3) At ~20 vol % melt (1000 ¿C), additional weakening occurs but fractures are not observed and the overall deformation is by viscous flow. The results show that at low melt fractions (<15 vol %) fluid-absent melting reactions can induce fracture in previously ductile rocks. This suggests that the percentage at which melt may escape by fracture is lower than the theoretical critical melt fraction or CMF of 26--40 vol %. In general, estimating the fraction of melt at the onset of segregation cannot be predicted by the CMF. Melt segregation models have to be adapted to variations in such major factors as pressure, temperature, type of melting reaction, rate of melting, and strain rate. Such models will greatly assist our understanding of continental growth and evolution during orogenesis. ¿ American Geophysical Union 1995