Defining the response of the human basilar membrane

Sound waves enter the ear canal and cause the eardrum to vibrate. These vibrations are transmitted through the middle ear, via three tiny bones (the ossicles), to the cochlea in the inner ear. The cochlea is a long thin tube, coiled up into a shape like a snail shell, that converts sound waves into electrical impulses in the nervous system. An important part of this process involves the basilar membrane, a thin membrane that runs the length of the cochlea. Each place on the basilar membrane is tuned to vibrate to a different frequency of sound, so that high frequencies (the bright crash of a cymbal) cause the membrane at the base of the cochlea to vibrate, and low frequencies (the dull thud of a bass drum) cause the membrane at the apex of the cochlea (the tip of the spiral) to vibrate. In this way the basilar membrane separates out the different frequency components of a sound, just as water droplets separate out the different frequencies of light (i.e. colours) to produce a rainbow. The frequency separation helps us to identify sounds, and to separate out sounds that occur together (for example, when listening to a conversation at a noisy party). The vibration of the basilar membrane is modified by special cells called outer hair cells. The outer hair cells amplify sounds at low levels but not at high levels, leading to a shallow growth of basilar membrane vibration with sound level called compression. This makes us much more sensitive to quiet sounds, without affecting our sensitivity to loud sounds. Furthermore, the amplification only affects a limited range of frequencies at each place on the basilar membrane. In this way, the amplification sharpens the tuning, making the membrane better at separating sounds. My research is concerned with measuring the effects of the outer hair cells on the vibration of the basilar membrane in humans. Although physiological experiments have been conducted on other mammals for a number of years, we have only recently developed the techniques necessary for accurate measurement of the human basilar membrane response. In these experiments, sounds are presented to participants over headphones and their responses are recorded (hence, this is a 'behavioural' experiment). Participants are asked to detect one sound (the 'signal') presented after another sound (the 'masker') with the same or a different frequency. By measuring how the detection of the signal depends on the levels of the signal and of the masker, it is possible to estimate the response of the basilar membrane without requiring a surgical procedure, a procedure that would be unethical in humans. As well as providing information about the human auditory system, the results will be relevant to mammalian hearing in general. For example, one of the aims of the research is to determine whether the basilar membrane is compressive near the apex of the cochlea spiral, the region that responds to low frequencies. Direct measurements of basilar membrane vibration in other mammals suggest that it is not, but these experiments may have been compromised because of the difficulties involved in the surgical procedure. It has been suggested that the outer hair cells were damaged, so that the amplification and compression were lost. The present behavioural experiments on humans should be able to resolve this issue because we can measure the cochlea in a healthy physiological state, and compare these results to those from hearing-impaired listeners who have damaged outer hair cells. We will use our results to develop a computational cochlear model, a computer programme that will allow us to simulate the response of the basilar membrane to any sound. Our improved understanding of how the basilar membrane works conveys information to more central structures will then help us understand how the brain uses that information to analyse and identify sounds.

The basilar membrane (BM) in the mammalian cochlea separates the frequency components of sounds before they are transduced into electrical activity in the auditory nerve. Each place on the membrane is tuned to a different 'characteristic' frequency (CF). In this way the spectrum of a sound is represented as a tonotopic spatial map on the BM and in the neural centres of the auditory system. Arranged in rows along the length of the BM are the outer hair cells (OHCs). These cells modify the response of the BM, amplifying low-level sound components with frequencies close to the CF of each place on the BM. In this way, the OHCs sharpen the tuning of the BM, enhancing the frequency selectivity of the auditory system as a whole. Because the amplification is lost at high levels, the growth in the level of vibration of the BM with sound level is shallow, i.e. compressive, with a compression exponent of about 0.2 (5:1 compression). Dysfunction of the OHCs is the main cause of sensorineural hearing loss, which is characterised by a reduction in frequency selectivity and a loss of compression (i.e. a linear BM response). Over the last ten years, we have developed behavioural techniques for estimating the response of the human BM. A participant is required to detect a pure-tone signal presented (over headphones) shortly after a masker sound: a design called 'forward masking'. In the 'temporal masking curve' (TMC) technique, the signal is fixed at a low level and the temporal gap between the masker and the signal is varied. For each gap, the level of the masker is found that just masks the signal. Because longer gaps require higher masker levels, the TMC allows us to measure masking as a function of masker level. Furthermore, because masker frequencies well below the signal frequency are not compressed at the signal place (at least at high CFs), a comparison of the TMCs for maskers at, and an octave below, the signal frequency can be used to estimate the on-frequency compression. A second technique is the 'additivity of forward masking' (AFM) technique. Two non-overlapping forward maskers are found that are equally effective at masking the signal when presented individually. When the maskers are combined, the signal detection threshold increases. The threshold increase depends on how much the signal is compressed. For example, under 5:1 compression the signal level has to increase by 15 dB to produce the 3 dB increase in BM vibration required to compensate for a doubling in masking effect. The proposed research uses these two techniques (and variations thereof) to answer three major outstanding questions regarding the response of the human BM: (i) How compressive is the BM at low CFs? (ii) How compressive is the BM to sound frequencies below CF? (iii) How compressive is the BM at low levels? These issues will be addressed in part by comparing the results from normal-hearing listeners to those from listeners with sensorineural hearing loss. Because of OHC damage, these listeners have reduced or absent compression, and they can be used as a control group to test hypotheses regarding normal hearing. To summarise, the research should supply the information required to complete our characterisation of the human BM. The results will be used to develop a computational model of the cochlea that will be a front end for models of hearing, and will enhance our understanding of the subsequent neural processing in the brainstem and in the auditory cortex. The research will also be relevant to our understanding of mammalian hearing in general, particularly for low CFs. At these frequencies direct physiological measurements of BM vibration may be compromised because of the difficulty of the surgical procedure. Our behavioural techniques will determine whether compression on the BM is reduced at low CFs, as suggested by the physiology, or the same as that at high CFs, as suggested by recent behavioural results.