This study assessed the test-retest reliability of distortion-product otoacoustic emissions (DPOAEs) at four frequencies (550, 1000, 2000 and 4000 Hz) over three time intervals. The time intervals were: (1) immediate test-retest reliability, in which the retest followed the test without any delay or repositioning of the probe tip; (2) very short-term test-retest reliability, in which the retest followed a 10-20-min break and involved removal and re-insertion of the probe tip; and (3) short-term test-retest reliability, in which the retest was conducted 5-10 days after the test. Fifty normal-hearing women were tested with a commercially available system for measuring DPOAEs (Grason-Stadler, GSI-60), which generated primary tones at 65 dB SPL (L1=L2). Standard errors of measurement at 550 Hz (approximately 4.6 dB) were nearly twice as large as those found for 1000 Hz, 2000 Hz, and 4000 Hz (approximately 2.5 dB). The short-term test-retest data suggest that there is a 95% probability that an individual's true DPOAE will fall within 5 dB of the obtained distortion product at 1000-4000 Hz and within 10 dB at 550 Hz. The standard error of measurement of the difference was calculated to assess whether two or more DPOAE measurements are significantly different (e.g. before versus after administration of an ototoxic drug or noise exposure). The data revealed that short-term differences (probe removed and subject retested on the same day or on different days) between two DPOAEs must exceed approximately 14 dB at 550 Hz and 7 dB at 1000-4000 Hz to be statistically significant at the 0.05 level of confidence.