Purpose: The purpose of this study was to examine test-retest reliability of low-level evoked distortion product otoacoustic emissions (DPOAEs) as a function of L[subscript 1], L[subscript 2] level; f[subscript 2] frequency; and test condition. A predictive relationship between these variables and the presence/absence of DPOAE responses was also examined. Method: Sixteen normal-hearing young adults participated. DPOAEs were evoked to 12 tones with f[subscript 2] frequencies ranging from 1500 Hz to 7546 Hz at 4 L[subscript 2] levels between 45 dB SPL and 30 dB SPL. Four test conditions were employed: (a) initial test, (b) retest without probe removal, (c) retest with probe reinsertion, and (d) retest with probe reinsertion by a second tester. Results: L[subscript 1], L[subscript 2] level and f[subscript 2] frequency were statistically significant (p less than 0.0001) predictors of a DPOAE response (i.e., the presence of a DPOAE response was more likely to be observed at higher L[subscript 1], L[subscript 2] levels and lower f[subscript 2] frequencies regardless of test condition). DPOAE levels were significantly affected by L[subscript 1], L[subscript 2] level and f[subscript 2] frequency (p less than 0.0001) but not by test condition. Intra- and intertester test-retest differences were not significantly different. Conclusions: The prevalence of missing responses coupled with large intersubject variability and intrasubject test-retest variability are a detriment to the clinical utility of DPOAEs evoked with low-level stimuli.