The feasibility of using ultrasonic techniques to detect fatigue damage in metal-matrix composites at an early stage has been investigated theoretically. The lack of direct independent experimental evidence for such damage required damage mechanisms and levels to be postulated in a somewhat arbitrary fashion. Estimates have been made of the effect of matrix porosity, microcrack formation, particle debonding and dislocation generation on ultrasonic velocity, attenuation and backscattering. It would seem that velocity measurements provide the best opportunity for detecting low levels of damage. Initial calculations suggest that surface wave velocity measurements and acoustoelastic constant measurements may also prove useful.