A bstractA mechanism is proposed for the anelastic relaxation e ect observed at high temperatures for Ni 3 A l by Gadaud and Chakib (1993, Mater. S ci. Forum , 119± 121, 397). The dependence of the relaxation magnitude on the crystal orientation as f ound by Mourisco et al. (1996, J. Phys. Paris, IV, C8-235) strongly suggests that the e ect is due to stress-induced redistribution of point defects on the sublattice sites for Ni atoms. Both the kinetic and the thermodynamic aspects of the relaxation can be explained quantitatively in terms of reorientation of antisite Al atoms. The properties of the point defects relevant to the e ect have been calculated by a thermodynamic model of the Bragg± Williams type and by atomistic computer simulation. The rate and the magnitude of the relaxation observed in experiment are in reasonable agreement with theoretical values estimated f rom the results of these calculations.