When a pair of tonal stimuli of different frequencies (F1 and F2, where F2 > F1) are simultaneously presented to the ear, an electrical response with a frequency of F2-F1 can be recorded from the round window (RW) of the gerbil's cochlea. By using phase-locked tones of alternating polarity, the cochlear microphonics are canceled, leaving a time-averaged difference tone-auditory nerve neurophonic (DT-ANN). When the F1 frequency ranges from 1.25 to 30 kHz and F2-F1 approximately 900 Hz, a DT-ANN audiogram can be constructed which parallels (but is at least 10 dB more sensitive than) the compound action potential (CAP) audiogram. In addition to this DT response, a smaller magnitude, higher threshold response having a frequency of 2 DT can often be measured. Both the DT-ANN and the 2 DT-ANN show non-monotonic amplitude input-output functions. The DT- and 2 DT-ANN responses can be forward masked. Masking of low level (e.g., 30 dB SPL) probe stimuli results in DT- and 2 DT-ANN V-shaped tuning curves (TC) with low tip thresholds (approximately 20-30 dB SPL) and a tip frequency close to that of F1 and F2. The Q10 dB values of the forward masked DT-ANN TCs ranges from 1.54 to 20.0 for F1 frequencies varying from 2 to 20 kHz, respectively. The V-shaped DT-ANN TCs generated with simultaneous maskers are often flanked, outside their high- and low-frequency slopes, by frequency-intensity domains where the masker enhances the amplitude of the DT-ANN response. These data (1) provide evidence that, in response to low-intensity tones, the DT-ANN is generated by a restricted population of neurons that have characteristic frequencies close to F1 and F2, and (2) provide evidence for sharply tuned, phase-locked activity occurring in response to low-intensity stimuli, by cochlear axons having characteristic frequencies as high as 20 kHz.