A method is proposed for studying robust stability of uncertain linear systems subject to actuator saturation. It hinges upon successive LMI relaxations and therefore can readily be implemented using widely available numerical tools. The technique is illustrated on two examples and is shown to significantly improve previously published results.