The GOM and TMC techniques have been used to measure the BM response in the base of the cochlea (CFs of about 2 kHz and above). The results are largely consistent with direct measures in other mammals using invasive techniques such as laser interferometry (Ruggero et al. 1997). The estimated response function at CF shows a linear growth for input levels up to about 30 dB SPL, followed by a highly compressive mid-level region extending from about 30 dB SPL to about 80 dB SPL. The compression exponent (slope of the response function) in this level region is typically in the range of 0.15–0.25 dB/dB, corresponding to a compression ratio of about 5:1 (Lopez-Poveda et al. 2003; Nelson et al. 2001; Oxenham and Plack 1997). Listeners with moderate-to-severe cochlear hearing loss show a loss of gain and compression and hence a linear response function at CF (Oxenham and Plack 1997).

It is known from physiological measures that compression in the apex is not frequency selective, and hence, the off-frequency masker cannot be used as a linear reference as it too will be compressed. This problem can be negotiated for the TMC technique by assuming that the rate of decay of forward masking in terms of BM excitation does not vary with frequency. The off-frequency TMC for a high signal frequency can then be used as a linear reference for all other frequencies (Plack and Drga 2003). For example, the 4-kHz off-frequency masker level can be plotted against the on (or off)-frequency masker level at the frequency of interest to derive a response function. Calculations such as this suggest that the basic shape of the BM response function in the apex is similar to that in the base, with a similar compression exponent (Lopez-Poveda et al. 2003; Nelson and Schroder 2004; Plack and Drga 2003). However, the range of levels that are compressed is smaller in the apex, and the range of frequencies relative to CF that are compressed is greater in the apex. Both of these findings are consistent with direct laser interferometry measures in other mammals. However, the amount of apical compression estimated in the human behavioural studies (about 5:1) is much greater than that measured directly in other mammals (about 2:1, Rhode and Cooper 1996). This could result from interspecies differences, although it is more likely a consequence of the difficulty of direct measures of the BM response in the apex, and the likelihood of damage to the active mechanism.

The additivity of forward masking (AFM) technique (Plack and O’Hanlon 2003) does not rely on a comparison of off- and on-frequency maskers. Instead, the technique is based on the assumption that the masking effect of two nonoverlapping forward maskers sums in a linear way. Hence, if two equally effective nonoverlapping forward maskers (M1 and M2) are combined, the internal excitation of the signal at masked threshold should increase by 3 dB (a doubling of intensity). However, if the signal is compressed prior to interacting with the maskers, the physical signal level will have to increase by more than 3 dB to produce an internal doubling in excitation. If the signal threshold in response to the individual maskers (M1 and M2) and the signal threshold in response to the combined maskers (M1  +  M2) are all known, it is possible to derive the compression exponent in the level region of the signal. The AFM technique can be used without substantial modification to measure the BM response at any CF. Results with the original AFM technique have largely confirmed the GOM and TMC results and in particular have confirmed the finding that compression in the base of the cochlea is as great as that in the apex and is similarly affected by cochlear damage (Plack et al. 2008). However, recent results from a new version of the AFM technique have raised the possibility that the AFM technique may be measuring more than just BM compression.

The estimates of gain and compression from the GOM and TMC techniques are thought to be largely unaffected by processes central to the BM. Any subsequent processing will affect on- and off-frequency maskers equally at the signal place, and there will be no differential effect. Theoretically, however, compression estimates from the AFM technique will be affected by any non-linear process prior to the neural interaction of maskers and signal. The correspondence of the earlier AFM results with the GOM and TMC results suggest that any compression central to the BM is relatively insignificant. However, recent results tell a different story. Plack and Arifianto (2010) used a variant of the AFM technique in which the signal is fixed at a low level (10-dB SL) and the masker-signal gap varied. Instead of varying signal level to determine the effect on threshold of combining two maskers, the signal level is fixed and the masker levels are varied to find the threshold values for each individual masker and the two maskers combined. By comparing masker levels at threshold for each masker alone and for the combined maskers, for each masker-signal interval, it is possible to infer the BM response function. Notice that the “masker-vary” AFM technique measures compression of the masker. Hence, it is possible to use this technique to measure off-frequency compression at the signal place by using masker frequencies different from the signal frequency.

Plack and Arifianto found that compression estimates at 4 kHz were roughly twice as great as reported previously, both on- and off-frequency (Fig. 4.2). Mid-level compression was about 10:1 for the CF response, and about 2:1 for the off-frequency response. Plack and Arifianto included conditions in which the maskers and signal covered a total duration of 40 ms: too short for substantial olivocochlear efferent activation at the time of presentation of the signal, which may affect compression estimates (Jennings et al. 2009). So how do we reconcile the earlier “signal-vary” AFM results with the more recent results? First, the earlier experiments used higher-level signals, rather than a fixed low-level signal. The maximum excitation on the BM produced by the signal moves to places with higher CFs as level is increased, places for which the signal frequency is below CF and hence is processed more linearly. In addition, previous studies used a longer-duration M1, which may produce efferent effects, reducing compression estimates (Plack and Arifianto 2010).