A patient's tolerance of fiberoptic bronchoscopy depends on the effectiveness of local anesthesia. This study compares the three different methods of local anesthesia in common use After sedation, patients (n = 53) received either 4 ml of 2.5 percent cocaine by intratracheal injection (TI) (n = 18), by bronchoscopic injection (BI) (n = 19), or had 4 ml of 4 percent lidocaine delivered by nebulizer 20 min before the procedure (NEB) (n = 16). Patients and bronchoscopists scored the procedure using visual analog (VAS) and severity scales. Objective measurements of cough counts and episodes of stridor were recorded by phonopneumography. Patients' VAS scores showed a clear preference for the transtracheal method compared with either bronchoscopically injected cocaine (p less than 0.001) or nebulized lidocaine (p less than 0.001). Patients also reported that the TI method produced less cough during intubation of the larynx and inspection of the airways (BI and NEB, p less than 0.01). The TI method was also preferred by the bronchoscopists (BI and NEB, p less than 0.001); they reported less cough and easier tracheal intubation. The mean cough count was significantly lower for the TI group, 49 (43) compared with 95 (52) for BI (p less than 0.01), and 81 (43) for the NEB group (p less than 0.05). Patients' and bronchoscopists' VAS showed significant correlation with cough (r = 0.63-69, p less than 0.01). Stridor occurred in only two patients after TI, compared with 15 in the other two groups. Extra local anesthesia was required by 16 patients after BI, by all the NEB group, but by only one patient after TI. Subjective and objective measurement shows that 4 ml of 2.5 percent cocaine injected into the trachea produced excellent local anesthesia for fiberoptic bronchoscopy, there were no extra complications, and it was the method preferred by both patients and bronchoscopists.