Carrara marble has been deformed experimentally at temperatures ranging between 500¿ and 1000 ¿C, at confining pressures of 200 and 300 MPa, and up to very large strains in extension and compression, in order to study the microstructural and mechanical property changes associated with dynamic recrystallization. Microstructural studies were made by optical microscopy on ultrathin sections, and automatic grain size measurement techniques were used. When the temperature is sufficiently high (≈600 ¿C) and the stresses are high enough for deformation twinning to occur, twin boundary migration is a powerful recrystallization mechanism that does not modify the grain size. At stress levels too low to activate twinning, recrystallization occurs in two stages: the formation of nuclei by grain boundary bulging and subgrain rotation recrystallization in the grain boundary regions, followed by a second stage of grain boundary migration recrystallization to a larger grain size that eventually overprints the entire rock volume. The recrystallization process requires larger prestrains at the lower temperatures.

The size to which migration-recrystallized grains grow seems to be limited by the stress and grain size dependent twinning field boundary or by the boundary between the dislocation creep and grain size sensitive flow mechanisms. Within the range of experimental observations, dynamically recrystallized grain size does not depend on strain rate or temperature. Separate empirical stress versus grain size relations are presented for rotation and migration recrystallization that may be used for palaeopiezometry. No unequivocal experimental evidence of weakening resulting from recrystallization to finer grain sizes at laboratory strain rates was found. However, extrapolation of the experimental data to low, natural strain rates suggests that weakening following recrystallization may occur in nature. ¿ American Geophysical Union 1995