The Eseka migmatites provide a good example of high- P—high- T anatexis of Archaean quartz-plagioclase-biotite grey gneisses. Partial melting resulted in the development of neosomes comprising a quartzo-feldspathic fraction of granodioritic to trondhjemitic composition and ferromagnesian minerals represented by hornblende mainly (15–45 vol.%), garnet in lesser amounts (4–8 vol.%) and scarce clinopyroxene. Thermobarometry suggests that anatexis occurred close to the amphibolite-granulite facies transition, under conditions estimated at around 9 kbar and 750°C. A late retrogression is documented by the growth of epidote and chlorite and by plagioclase albitization. Textural relations, mineralogical data and recalculation of the primary mineral assemblages suggest that the biotite dehydration melting conditions were reached leading to the generation of granodioritic melts in equilibrium with K-feldspar, hornblende and garnet (±clinopyroxene). In some cases, this was followed, during melt crystallization, by a retrograde reaction involving the granodioritic melt and hornblende to produce biotite-epidote assemblages and trondhjemitic neosomes. Compared to similar migmatites, the Eseka neosomes are characterized by higher Al and Ti contents of hornblende, the presence of garnet and the lack of sphene which can be related to higher P, T conditions. Geochemical data point to the segregation and removal of some melt in the Eseka migmatites resulting in limited concentration of hornblende and garnet at the site of melting. This together with previous field observations and experimental data, could suggest that some hornblende-rich gneisses or amphibolites found in the lower middle crust may represent a residue left after extraction of anatectic melts. Geochronological and petrological data further suggest that large portions of the Archaean TTG rocks from the Congo craton have been reworked under high-grade conditions during the Paleoproterozoic.