Purpose: The gaps-in-noise (GIN) test was developed as a clinical evaluation of temporal resolution. In the GIN test, a sensation level of 50 dB SL was recommended as a standard clinical presentation level to ensure stimulus audibility. It has been reported that the stimulus level less than 30 dB SL should not be used as presentation level to verify a proper level for administration of GIN test. Given that the previous finding was based on the data of young normal-hearing adults, the present study aimed to compare the GIN results as a function of stimulus level in listener groups differing in age and hearing sensitivity. Methods: Participants were 10 young normal-hearing (YNH), 12 elderly normal-hearing, and 12 elderly hearing-impaired (EHI) listeners. GIN was administered at 20, 30, 40, and 50 dB SL. The performance was determined from both the approximate threshold and the percentage of correct gap. Results: As results, both approximate threshold and percent correct in GIN results were measured. Results showed that, overall, both the threshold as well as the percent correct in GIN improved with increasing stimulus level. Across groups, EHI showed the poorest results from GIN test, with the best performance in YNH. EHI listeners with greater hearing loss at higher frequencies (2-8 kHz) were poorer at detecting gaps in noise. Across a range of stimulus level (20-50 dB SL), the GIN performance between 40 and 50 dB SL did not differ in all listener groups. Conclusion: A sensation level of 35-50 dB may be applicable to the elderly with normal hearing or slight-to-mild degree of hearing loss, without significant change in GIN performance.