Effects of sample size on the latency and amplitude of the auditory evoked response.

Experiment I investigated the effects of sample size (500 to 1500 stimulus repetitions) on the auditory brainstem response as a function of intensity (20 to 80 dB nHL) on a group of 10 normally hearing subjects. There was little change in identifiability, reliability, latencies, or amplitudes of Waves I, III, and V as the sample size increased from 500 or 750 to 1500 repetitions. These results suggest that 500 to 750 repetitions may be adequate when methods similar to those in the present study are used, and that the common clinical practice of employing approximately 1500 repetitions may unnecessarily prolong testing. Experiment II employed 12 hearing-impaired subjects who were tested at 10 to 40 dB SL using sample sizes from 250 to 4500 stimulus repetitions. Identifiability of all waves increased as sample size increased from 250 to 4500 repetitions. The largest changes in identifiability occurred when sample size increased from 250 to 1500 or 3000 repetitions, with little improvement as sample size increased from 3000 to 4500 repetitions. Examiners should monitor averaged responses and terminate testing as soon as a wave is identified. Contrary to expectation, there was no systematic change in the standard error of measurement for latency (approximately 0.07 ms) as sample size increased from 250 to 4500 repetitions. The standard error of measurement for amplitude decreased from approximately 100 nV with 500 repetitions to approximately 45 nV at 3000 repetitions. The improvement in reliability with increasing sample size may be explained by a decrease in the variability of background noise. A systematic decrease in amplitude also was observed as sample size increased. This observation may be explained by a reduction in the residual noise levels or because of time jitter or adaptation within the auditory pathways. Nonetheless, investigators who wish to use ABR amplitude measures for diagnosis may benefit from using a relatively large sample size.