Objective: To study the incidence of phrenic neuropathy following coronary artery bypass grafting and determine long-term outcome. Design: Prospective observational. Setting: Surgical ICU in a university hospital, out-patient follow-up. Patients: Ninety-two consecutive patients undergoing open heart surgery. Interventions: None. Measurements: Chest radiographs (CXR) 48 to 72 h post-operatively, ultrasonography of diaphragm, phrenic nerve conduction studies, diaphragmatic electromyogram, each repeated every 1 to 3 months until normal. Main results: Seventy-eight of 92 (78 percent) patients had abnormal radiographs, 42 of 78 (54 percent) with abnormal CXRs had abnormal diaphragm motion, 24 of 42 (57 percent) with abnormal motion had phrenic neuropathy. Patients with normal diaphragm motion improved faster than those without; patients with normal nerve conduction (and abnormal motion) improved faster than those with abnormal nerve conduction. Conclusions: Phrenic neuropathy is relatively common if sensitive tests are utilized for diagnosis. Nerve conduction studies can predict duration of morbidity. Most patients have low morbidity and recover fully. Abnormal diaphragm motion alone is not diagnostic of phrenic nerve injury.