Although physiological measures have provided a great deal of -information about the basilar membrane (BM) response of non-human mammals, it is only relatively recently that behavioural techniques have allowed researchers to measure accurately the non-linear characteristics of the human BM. These techniques are based on forward masking, in which the threshold for detecting a signal is measured in the presence of a prior masking sound. Two popular techniques, the growth of forward masking technique and the temporal masking curve technique, rely on the fact that compression in the base of the cochlea is largely restricted to frequencies close to the characteristic frequency (CF) of each place. By comparing the response to a masker with a frequency equal to that of the signal with the response to a lower-frequency masker, it is possible to infer the CF response. These measures have shown that BM compression in humans matches that of other mammals and that compression is absent in listeners with moderate-to-severe cochlear hearing loss, probably reflecting outer hair cell dysfunction. Another technique, the additivity of forward masking (AFM) technique, does not rely on a comparison between on- and off-frequency maskers, but instead measures the effect on threshold of combining two nonoverlapping maskers, an effect which is magnified by compression. The difference between thresholds in the single- and combined-masker conditions can be used to estimate compression. The AFM technique has provided evidence that strong compression extends down to low CFs in humans, a finding inconsistent with direct measures of the BM response in other mammals. Furthermore, recent AFM results suggest that there may be an additional source of compression central to the BM. This more central compression also appears to be affected by hearing loss and may reflect non-linear processes in the transduction mechanism of the inner hair cells.