From ear canal to cochlea

Sound entering the ear travels through the canal, vibrates the eardrum, and is carried by three small bones to the cochlea — the coiled, fluid-filled organ of hearing. Inside the cochlea sit two kinds of sensory cells: inner hair cells, which send hearing signals to the brain, and outer hair cells, which act as tiny biological amplifiers.

Otoacoustic emissions come from those outer hair cells. When they are healthy, they actively move — and that movement produces a faint sound that travels back out of the ear. Recording that sound tells us the outer hair cells are working.

The cochlear amplifier

Outer hair cells are electromotile: they change length in response to changes in membrane voltage, driven by the motor protein prestin in their lateral wall[4]. This length change feeds mechanical energy back into the travelling wave on the basilar membrane on a cycle-by-cycle basis, sharpening the wave's peak and boosting low-level sounds by 40–60 dB. The by-product of this active process — energy radiating backward through the middle ear — is the otoacoustic emission [1].

The cochlea is tonotopic: the stiff basal end responds to high frequencies, the floppy apical end to low frequencies. Because of this, the high-frequency components of a transient emission return to the ear canal first, and the low-frequency components arrive later — the latency dispersion that distributes each frequency to its own characteristic place along the membrane.

Two physical mechanisms generate the emissions we record, and they map onto how each emission is evoked. Linear reflection from fixed irregularities along the cochlear partition produces the place-fixed component, while a nonlinear distortion mechanism in the region of overlap between two stimulus tones produces a wave-fixed component [2]. This dual origin explains why distortion-product and reflection-based emissions can behave differently even in the same ear [3].

The medial olivocochlear efferent system

Outer hair cell gain is not fixed. The medial olivocochlear (MOC) bundle projects from the brainstem onto the outer hair cells and, when activated — for example by contralateral noise — reduces their gain. This efferent suppression measurably lowers emission amplitude and is itself a probe of brainstem-level auditory function[7].

Clinically, the key consequence of this anatomy: emissions test the pre-neural cochlea only. A normal emission with an abnormal auditory brainstem response is the signature of auditory neuropathy spectrum disorder — outer hair cells intact, neural transmission impaired [6]. This dissociation is the reason emissions and the brainstem response are read together rather than in isolation.

Spontaneous emissions (SOAEs)

Not all otoacoustic emissions are the same. They are sorted first by a simple question: did the ear need a sound to produce the emission, or did it produce one on its own?

Some healthy ears produce a faint, continuous tone with no stimulus at all — a spontaneous otoacoustic emission. They are a normal finding, not a sign of disease. Occasionally a person can even perceive their own SOAE as a soft ringing[1].

Evoked emissions (EOAEs)

Far more useful in the clinic are evoked emissions — the ear's response when we deliberately play a sound into it. The two we record routinely are the transient-evoked OAE (TEOAE), the ear's echo after a brief click, and the distortion-product OAE (DPOAE), produced when two tones are played together.

The four classical types

The conventional classification recognises four emission types. SOAEs occur without stimulation. The three evoked types differ by stimulus: a TEOAE follows a brief click or tone-burst; a stimulus-frequency OAE (SFOAE) is the response to a single sustained pure tone; and a DPOAE is evoked by two simultaneous tones, the primaries f₁ and f₂. SFOAEs are difficult to separate from the stimulus itself and so see little routine clinical use, while TEOAEs and DPOAEs dominate practice[5].

The DPOAE is the most elegant of the four to measure, because the ear generates energy at a frequency that was never presented. With two primaries close together, the largest distortion product in the human ear appears at the frequency 2f₁−f₂. Because that frequency differs from both stimulus tones, the emission is easy to isolate from the stimulus in the recording.

A historical note worth keeping: David Kemp first demonstrated in 1978 that the cochlea emits measurable sound energy back into the ear canal, overturning the view of the ear as a purely passive receiver[1]. The discovery that isolated outer hair cells physically change length — electromotility — supplied the mechanism a few years later[4].

The dual-source model

The clinically important reframing is not the four-type list but the mechanism-based taxonomy of Shera and Guinan. Rather than every emission arising from a single nonlinear process, evoked emissions are generated by two fundamentally different intracochlear mechanisms: nonlinear distortion and linear coherent reflection[2].

A distortion-source emission is created where the traveling waves of the stimulus tones overlap and the cochlear response saturates. Because that overlap region moves when stimulus frequency changes, this is termed a wave-fixed source. A reflection-source emission instead arises from coherent backscatter of the traveling wave off fixed micromechanical irregularities distributed along the cochlear partition — a place-fixed source[2].

This matters clinically because the DPOAE recorded at 2f₁−f₂ is not a pure distortion-source emission: it is the vector sum of a distortion component generated near the f₂ place and a reflection component from the 2f₁−f₂ characteristic place. Their interference produces the fine structure seen in a DP-gram, and the two components can carry non-redundant information about cochlear health and may differ in their sensitivity to pathology and to ageing[3][7].

The practical takeaway: the TEOAE and click-evoked emission behave predominantly as reflection-source emissions, while the DPOAE is a mixed measure. Awareness of the dual-source model explains why DP-gram fine structure exists, why component-separation techniques are used in research protocols, and why no single emission type is a complete probe of cochlear function. For full treatment, see the standard text by Dhar and Hall[6].

Getting a good seal

Recording an otoacoustic emission needs only one piece of hardware in the ear: a small probe. Inside it sit a tiny loudspeaker, which plays the stimulus, and a sensitive microphone, which listens for the faint sound the cochlea sends back.

The probe is fitted with a soft foam or rubber tip that must seal the ear canal completely. The seal matters more than almost anything else in the test: if it leaks, the stimulus escapes and the quiet emission is lost in room noise. A poor seal is the single most common reason an otherwise healthy ear produces a “refer” result[8].

The test environment should be quiet, and the patient still and relaxed — ideally asleep, in the case of a newborn. Sucking, crying, and movement all add noise that the instrument must reject before it can find the emission.

The TEOAE stimulus

A transient-evoked OAE is elicited with a brief broadband click. In routine practice the click is presented in a non-linear paradigm — a train in which one click is inverted and scaled — which cancels the linear stimulus artifact and leaves the non-linear cochlear response. Non-linear click levels are typically set around 80–84 dB peak-equivalent SPL; the lower-level linear click is used less often because it is more prone to artifact[8].

The response is recorded in a short time window — on the order of 20 ms — with the first few milliseconds blanked to exclude stimulus ringing. Because a single click's emission is buried in noise, the instrument averages many sweeps: the random noise cancels over repeated trials while the repeatable emission builds up.

The DPOAE stimulus

A distortion-product OAE is evoked by two pure tones presented together — the primaries f₁ and f₂, with f₂ > f₁. Their spacing and level are not arbitrary; the table below lists the parameters that routine protocols converge on.

Values reflect commonly recommended protocols[8]; exact settings vary by device and clinical purpose.

Quality control during recording

A valid emission depends on three things being right at once, and a careful clinician watches all of them. Stimulus stability should stay high throughout the run — a stability figure that drifts well below the high-90s percent means the probe moved or the seal changed, and the measurement should be repeated. Noise rejection discards individual sweeps that exceed a set level so they never enter the average; turning it off to save time degrades the data and is not advised. And the stimulus spectrum must be genuinely broadband — a click missing high- or low-frequency energy simply cannot evoke a response from the corresponding region of the cochlea[8].

The result is read as a signal-to-noise ratio: the emission must rise a criterion margin above the noise floor — commonly around 6 dB — and be reproducible. Screening protocols typically require a pass in a set number of frequency bands, for example three of four[9].

Common technical pitfalls

Before attributing an absent emission to cochlear pathology, exclude the avoidable causes. Cerumen or debris occluding the probe tip, vernix in a newborn's canal, a partial seal, a probe pressed against the canal wall, and a noisy or unsettled patient all produce a refer result in an ear that may hear perfectly well. Middle-ear effusion does the same — which is why probe checks, otoscopy, and tympanometry belong alongside the OAE, not after it[6].

A normal TEOAE

The two recordings you will see most often are the TEOAE response waveform and the DPOAE “DP-gram”. After a click, a healthy ear returns a brief, oscillating emission in the first 20 milliseconds or so. It is small, but it is real and it repeats: record it twice and the two traces look alike.

A normal DP-gram

The DP-gram plots the emission level against test frequency. In a normal ear every point sits comfortably above the noise floor — the shaded band beneath it.

When is an emission “present”?

A response is judged present when two conditions are met together: it must exceed the noise floor by a criterion signal-to-noise ratio, and it must be reproducible. For TEOAEs an acceptable response is commonly described as roughly 3–6 dB above the noise floor; if a reproducible response is not seen at an SNR of at least 3 dB, the outer hair cells of that frequency region cannot be assumed to be functioning normally[10].

Reproducibility is quantified as the correlation between two independently averaged buffers — often labelled A and B. A high waveform-reproducibility percentage, close to 100%, indicates the two buffers agree and the recorded response is genuine rather than noise[9].

How OAEs map onto the audiogram

Normal emissions correspond to good pure-tone hearing, within limits. TEOAEs are present in about 99% of ears when all pure-tone thresholds are better than 20 dB HL, and are essentially always absent once thresholds exceed 40 dB HL; between roughly 25 and 35 dB HL they may or may not appear. DPOAEs behave similarly, though with stronger primaries a reduced DPOAE can sometimes persist with losses up to 50–60 dB HL[10].

Figures summarise widely cited clinical data[10]; thresholds are approximate and depend on protocol.

Present, present-but-abnormal, absent

Screening reduces the result to pass or refer, but diagnostic interpretation uses three categories. An emission that meets the SNR criterion and falls within the normative amplitude range is present and normal. One that meets the SNR criterion but whose amplitude lies below the normal range is present but abnormal — a meaningful finding that a pass/refer screen would miss. One that fails the SNR criterion is absent. This three-way classification, anchored to frequency-specific normative data, is the basis of the landmark large-scale DPOAE studies[10].

Normal variation: age, sex, and SOAEs

“Normal” is a range, not a single value. Emission amplitudes are notably larger in infants than in adults — by several decibels, more so at high frequencies — which is part of why OAEs work so well for newborn screening. Amplitudes tend to be modestly larger in females than in males, and they decline gradually with age, an effect partly but not entirely explained by accompanying threshold change[11].

The presence of spontaneous emissions is itself a marker of a robust cochlea: ears with SOAEs tend to show larger evoked-emission amplitudes and SNRs. Because of all this variation, the strongest normative comparison is frequency-specific and, ideally, age-appropriate — adult templates can misclassify infant ears, whose emissions are genuinely larger[10].

What a result actually tells you

A present emission means the outer hair cells in the tested region are working and the sound path to and from them is clear. An absent emission means something along that chain is not working — but it does not say what. The simple rule: if you can record an emission the inner ear is working; if you cannot, it does not necessarily mean the inner ear is broken[8].

And an emission does not measure how much someone hears. A pass means hearing is probably better than about 30–40 dB HL in that region — it is not an audiogram, and it puts no number on the threshold[10].

The false refer

OAE screening is sensitive but not very specific: a refer result in a normal-hearing ear is common. Before attributing an absent emission to the cochlea, work through the avoidable causes — cerumen or debris in the canal, vernix in a newborn, a partial probe seal, the probe resting against the canal wall, an unsettled or noisy patient, and background noise in the room. Internal sounds — breathing, swallowing, sucking — contaminate the recording just as external noise does[8].

Middle-ear status is the single most important confounder. Because the emission must travel out through the middle ear, any middle-ear problem can abolish it even when the cochlea is perfectly healthy. Effusion, negative pressure, or adhesive otitis media all produce a refer result indistinguishable from a cochlear one — the OAE cannot, by itself, separate conductive from sensorineural loss[6].

Trust the instrument first

A spurious result sometimes comes from the equipment, not the patient. A daily probe-cavity check — running the probe in a test cavity or ear simulator — confirms the instrument reaches its noise floor and shows no false emissions above it. If a signal does appear, suspect debris or wax in the probe before testing anyone[8].

The cross-check principle

No single test should stand alone. The cross-check principle, set out by Jerger and Hayes in 1976, holds that the result of any one test must be confirmed by an independent measure before it drives a diagnosis. For the OAE, the indispensable cross-checks are tympanometry — which settles the middle-ear question — and the auditory brainstem response, which probes the neural pathway the OAE cannot see[12].

Reading OAE against ABR

Because the OAE tests the pre-neural cochlea and the ABR tests the neural pathway, the two together resolve cases either test alone leaves ambiguous. The matrix below is the core interpretation aid.

The clinically dangerous quadrant is present OAE with abnormal ABR: hearing can be seriously impaired while the emission looks entirely normal. This is auditory neuropathy spectrum disorder, and it is the reason an OAE-only newborn screen is unsafe — an automated ABR is needed to catch it[8].

Beyond pass/refer

Screening collapses the result to two outcomes, but diagnostic interpretation uses three: present and within the normal amplitude range; present but abnormal — meeting the signal-to-noise criterion yet below the normative range; and absent. The middle category carries real information about early or partial outer hair cell dysfunction that a pass/refer screen discards entirely. Diagnostic reading should always be frequency-specific and compared against appropriate, ideally age-matched, normative data[10].

Three rules to carry

First, a refer is not a diagnosis — exclude debris, seal, noise, and middle ear before the cochlea. Second, a pass is not a clean bill of hearing — it cannot exclude auditory neuropathy or a loss milder than the test's floor. Third, interpret the OAE inside a battery — tympanometry and ABR are not optional extras but the cross-checks that make the OAE safe to act on[12].

The 1-3-6 benchmark

Otoacoustic emissions found their largest role here: screening the hearing of newborn babies. The test is quick, needs no response from the baby, and can be done while the infant sleeps — which is why OAE screening has been adopted for newborns across most of the world.

The reason it matters is timing. A baby whose hearing loss is found and treated early develops markedly better language than one identified late. Screening exists to close that gap — to catch hearing loss in the first weeks of life rather than the first years[13].

Early hearing detection and intervention programmes are organised around three deadlines, known together as 1-3-6.

The two-stage screen

Screening is not a single test but a pathway. An initial screen is performed — often only after the first 12 hours of life, so birth fluid and debris can clear the ear canal. A baby who does not pass is rescreened rather than immediately labelled. Only a baby who fails the rescreen is referred onward for full diagnostic assessment.

Two principles govern the pathway. A true pass requires both ears to pass within the same session — a one-ear pass is not a pass. And rescreening is deliberately limited: excessive re-testing raises the chance of wrongly passing a baby who does have hearing loss, so a persistent refer should move to diagnostic audiology, not into another round of screening.

Programmes aim to keep the refer rate low — commonly a target below about 4% — because every false refer means an anxious family and an unnecessary appointment[13].

OAE or automated ABR?

Two technologies screen newborns: OAE and automated auditory brainstem response (AABR). They test different parts of the pathway, and the difference is decisive. The OAE response is generated by the outer hair cells before the signal reaches the eighth nerve — a pre-neural response. AABR follows the signal onward along the nerve to the brainstem[13].

This is why an OAE-only screen has a blind spot. An infant with auditory neuropathy spectrum disorder has working outer hair cells and so passes an OAE screen — while AABR, which tests the neural pathway, correctly refers. An infant with ANSD screened by OAE alone is passed and lost to follow-up.

Why the NICU rule exists

Infants cared for in a neonatal intensive care unit are at substantially higher risk of neural hearing loss, including ANSD. For that reason many programmes mandate AABR — not OAE — as the screening method for NICU infants. The OAE simply cannot evaluate the auditory nerve or brainstem, so in the population most likely to have a neural lesion it is the wrong tool. A refer on an AABR screen should also not be “rescreened” with an OAE, since that would substitute a test that cannot see the suspected problem[13].

The combined approach

The strongest protocols use both tests rather than choosing between them — an application of the cross-check principle to screening. OAE screening is efficient and effective for detecting the cochlear and transient middle-ear losses common in well babies; AABR adds coverage of the neural pathway. Combining them lowers the false-refer rate and ensures auditory neuropathy is not missed[12].

Whichever protocol is used, screening is only the first step. A baby referred from screening still needs a full diagnostic evaluation — the screen says “look closer here”, not “this is the diagnosis”.