The cancellation level and phase were measured for the (f2−f1) distortion product in six normal‐hearing ears as a function of input level (L1,L2) and the frequency separation of the two input signals (f1, f2, where f2>f1). The effects of input level were examined for L1 and L2, varied together (L1=L2) and separately. Typically, f1 was 1500 Hz while f2/f1 was either 1.16, 1.32, 1.44, or 1.68. For L1=L2, the median data indicate that the (f2−f1) level grows at a rate of approximately 1.1 dB/dB when averaged across all stimulus conditions. This slope tended to be higher (∼1.6 dB/dB) for L1=L2>80 dB. Slopes for some subjects also increase as f2/f1 increases. The cancellation phase increases slightly (50°–100°) with an increase in L1=L2. With L1 at 75 to 80 dB, L2 ranging from 65 to 95 dB, and f2/f1=1.16, (f2−f1) increases monotonically with L2 up to L1=L2. As L2 increases further (L2>L1), the slopes for the growth of (f2−f1) decrease. For f2/f1=1.44, on the other hand, (f2−f1) appears to grow monotonically at a rate of approximately 0.5 dB/dB throughout the range of L2 values. The cancellation phase increases with L2 (approximately 100°) only for the wider frequency separation of the two input signals (f2/f1=1.44). There are, however, sizable individual differences in the behavior of the (f2−f1) distortion product.