Fourteen rested and fatigued residents read a set of 25 test radiographs on two separate occasions approximately one month apart. The radiographs were made with partial chest phantom using pulmonary nodules as test objects. Seven rested and 7 fatigued residents read the radiographs, and a month later the order of fatigue was reversed. Fatigued residents had worked a minimum of 15 consecutive hours before their interpretations. Performance did not deteriorate with fatigue, and the nodule detection rate was almost identical on both occasions.