5.1 The cochlear amplifier as clinical signal source

The basilar membrane in a passive model of the cochlea is a tonotopically tuned mechanical bandpass filter. A puff of sound enters at the oval window, sets up a traveling wave on the membrane, and that wave peaks at the place whose mechanical properties match the input frequency. The peak excites the inner hair cells at that place, which fire afferent fibres in the auditory nerve, and the brain receives a frequency-organised neural signal. The classical von Békésy picture, developed in the 1930s–1950s, treated the cochlea this way and won Békésy the 1961 Nobel Prize.

The picture is right but incomplete in one critical respect: the passive basilar membrane, as Békésy measured it in cadaver preparations, has a peak quality factor of about $Q \approx 2$ — far too broad to explain the frequency discrimination thresholds (about 0.2% at 1 kHz, equivalent to $Q \approx 500$ ) that human listeners achieve. Something living and active is sharpening the response in the live cochlea. That something is the cochlear amplifier, and its existence in mammals was inferred from physiological measurements in the 1970s but became directly observable with Kemp’s 1978 discovery of the otoacoustic emission. The OAE is the cochlear amplifier’s exhaust.

What outer hair cells do

The two classes of cochlear hair cell have distinct roles. The inner hair cells (IHCs, about 3,500 in a human cochlea, arranged in a single row) are the sensors: their stereocilia bend with basilar-membrane motion, mechanotransduction channels open, the cell depolarises, and the afferent dendrites of the auditory nerve fire spikes encoding sound. The outer hair cells (OHCs, about 12,000, three to five rows depending on position) are not primarily sensors. Their afferent innervation is sparse and they make only a minor direct contribution to the neural signal that reaches the brain. Their role is mechanical: they shorten and lengthen in response to changes in their membrane potential, on a cycle-by-cycle basis up to at least 70 kHz, driven by a unique motor protein called prestin packed into their lateral cell membrane. Discovered by Peter Dallos’s lab in 2000, prestin is a piezoelectric-like transmembrane protein that converts changes in transmembrane voltage directly into changes in cell length, at speeds and amplitudes no other known biological motor approaches.

The mechanical loop is: basilar membrane moves → OHC stereocilia bend → mechanotransduction channels open → OHC depolarises → prestin contracts → OHC shortens → tectorial membrane is pulled toward the OHC → basilar membrane is further deflected. This positive-feedback loop, operating in a narrow frequency band around each OHC’s characteristic frequency, adds energy to the traveling wave at the peak of its envelope — sharpening the tuning, lowering the threshold by 40–60 dB, and producing the steep compressive nonlinearity that makes the cochlea the working organ that it is.

Why the OAE exists

The cochlear amplifier produces more mechanical energy at the peak than the input wave delivered to that place. That energy doesn’t only go forward into the inner hair cells; some of it propagates backward as a traveling wave on the basilar membrane, eventually reaching the stapes, vibrating the ossicular chain in reverse, and pressing on the tympanic membrane. The middle ear, which is a nearly-reciprocal transmission line, allows the reverse signal to escape as sound in the ear canal. That escaping sound is the otoacoustic emission.

In a normal cochlea, OAEs occur:

Spontaneously (SOAE) in about 50% of normal-hearing ears — narrow-band tones at one or a few frequencies, audible if amplified by a low-noise probe, generated by the cochlea idling at a stable limit cycle near its operating point. SOAEs are clinical curiosities and rarely tested; the ones that are loud enough to be perceived appear as a form of tinnitus.
Transient-evoked (TEOAE) when stimulated by a click — covered in Lesson 5.2.
Distortion-product (DPOAE) when stimulated by two simultaneous primary tones — covered in Lesson 5.3.
Stimulus-frequency (SFOAE) when stimulated by a long pure tone — measured in research labs but not routinely clinical.

The three evoked types differ in how the cochlea is excited, but they all originate in the same physical process: the active mechanical contribution of the outer hair cells.

What the OAE tells us clinically

Two facts make the OAE clinically powerful:

It is highly specific to OHC function. Inner-hair-cell loss without OHC loss is rare but possible (auditory neuropathy spectrum disorder, ANSD); these patients have normal OAEs despite having sensorineural hearing loss. Conversely, the most common form of acquired sensorineural loss — noise damage, ototoxicity, age-related — disproportionately damages OHCs first, and the OAE typically vanishes before the audiometric threshold has clinically risen. The OAE is therefore a sensitive early warning sign for cochlear damage that has not yet produced behavioural loss.
It requires no behavioural cooperation. This is what makes universal newborn hearing screening possible. The infant lies asleep in the bassinet; the probe is placed in the ear; the test takes 30 seconds per ear. Confirmation testing in referred infants uses automated auditory brainstem response (AABR, Ch 6) — also behavioural-cooperation-free.

OAE testing also has clear limitations:

OAEs vanish above ~30–40 dB HL of cochlear loss. They are present-or-absent indicators of cochlear health, not graded indicators of degree of loss.
OAEs are affected by middle-ear status. A type-B tympanogram (effusion) prevents the OAE from reaching the probe even if the cochlea is healthy. Newborn screens routinely fail-because-of-middle-ear-fluid; this is why protocols include AABR backup.
OAEs say nothing about retrocochlear or central pathology. A patient with a vestibular schwannoma but intact OHCs will have normal OAEs.

⏳ The history — Kemp's 1978 discovery

David Kemp, then a graduate student at the Institute of Laryngology and Otology in London, was studying ear-canal acoustics in 1977. He noticed that when he presented a click stimulus, his microphone recorded — buried in averaging — an echo arriving about 10 ms after the click. The echo’s amplitude depended nonlinearly on the click level (it saturated at high stimulus levels but was clearly present at low ones), and its spectral content showed the click had been broken down by frequency: high components arriving first, low components arriving later. The signal was unmistakably from inside the ear, not a microphone artefact, but no passive structure could account for the energy.

Kemp’s 1978 paper (Stimulated acoustic emissions from within the human auditory system, JASA 64: 1386–1391) proposed that the cochlea contains an active mechanical element that does work on the traveling wave. The proposal was controversial — Békésy’s passive cochlea was the dominant picture, and adding active elements seemed to conflict with the steady-state energy budget — but the evidence rapidly accumulated. Mountain (1980) showed efferent activation of the medial olivocochlear bundle could suppress OAEs; Brownell et al. (1985) discovered that isolated OHCs in tissue culture changed length in response to current injection; the prestin gene was cloned by Dallos’s group in 2000.

The first clinical OAE system (the Otodynamics ILO88) was launched by Kemp’s own company in 1988. By the late 1990s, multiple US states had universal-newborn-screening mandates that effectively required OAE technology in every birthing hospital. Kemp’s discovery moved from curiosity to standard-of-care in 25 years — a faster trajectory than most basic-science findings achieve.

The next two lessons cover the two evoked OAE types audiologists actually use clinically: the transient-evoked OAE (a single-click screening tool) and the distortion-product OAE (a frequency-specific diagnostic tool that produces a clinical analogue of the audiogram).