Respiratory arrest is a serious complication of retrobulbar anesthesia. We have prospectively followed 3123 retrobulbar injections to determine the incidence of respiratory arrest and identify possible risk factors. Injections contained either 2% or 4% lidocaine with 0.75% bupivacaine (Marcaine) in a 50:50 mixture by volume (total 10 cc.) to which a 1 ml ampule of hyaluronidase (150 NF units) was added. Use of 4% lidocaine gave a significantly higher incidence of respiratory arrest than 2% lidocaine (0.79% vs. 0.09%; P = 0.003, Fisher's exact test). Serum levels of lidocaine and bupivacaine of patients who arrested were not elevated significantly compared to 20 control patients. All levels were well below accepted levels of toxicity. Thus, 4% lidocaine increases the risk of respiratory arrest. These results suggest the mechanism is not direct intravascular or central nervous system injection.