One pathway, many responses

A single sound sets off electrical activity all the way up the ascending auditory pathway — and we can record it from the scalp at every level, separated by whenit arrives. Within the first 10 milliseconds we see the auditory brainstem response (ABR) from the eighth nerve and brainstem nuclei. Between roughly 10 and 50 ms the middle latency response appears, with contributions from the thalamus and primary cortex. From about 50 ms onward comes the slow, late cortical auditory evoked potential— the P1–N1–P2 complex [3].

Each response interrogates a different stretch of the same pathway. The ABR asks whether the nerve and brainstem can fire in tight synchrony; the cortical response asks a higher-level question — whether sound has been detected and processed at the cortex. That difference is the whole clinical value of the late response.

Where the late response is generated

The obligatory cortical response is not generated at a single point. Its largest component, N1, has several overlapping generators in and around the supratemporal plane of the auditory cortex, together with contributions that give it its characteristic vertex-maximum scalp distribution [4]. P1 and P2 reflect activity in primary and secondary auditory cortical areas and their thalamocortical input. Because the sources sit in auditory cortex, the response is largest at the top of the head — the reason the recording electrode goes at the vertex.

The decisive clinical fact follows directly from this anatomy: a present cortical response means the signal travelled the entire pathway and engaged the cortex. The slow vertex response to sound was described in the earliest days of human electroencephalography[1], and its dependence on stimulus level — the property that makes it useful for audiometry — was quantified soon after [2].

Obligatory versus cognitive responses

The cortical responses split into two families. The obligatory (or exogenous) responses — P1, N1 and P2 — depend on the physical stimulus and appear whether or not the listener attends to the sound. This is exactly what makes them usable in the clinic: a child or a medicolegal claimant cannot suppress them at will, and they can be recorded passively while the patient sits quietly [5].

The cognitive (endogenous) responses — the P300 and the mismatch negativity— depend on attention, memory and the detection of an oddball among standard sounds. They index processing rather than audibility, are far more variable, and are largely research tools. Throughout this atlas, “the cortical response” means the obligatory P1–N1–P2 complex unless stated otherwise.

A cortex that must be awake

One more consequence of generating the response in the cortex: it depends on arousal. The ABR is generated low in the pathway and survives sleep and sedation, which is why infants are often tested asleep. The late cortical response is the opposite — N1 in particular shrinks and degrades in drowsiness and sleep[5]. Cortical audiometry therefore requires an awake, relaxed but alert patient, a constraint that shapes the whole recording technique.

Reading the waveform

By convention the components are named by polarity and order — P for a vertex-positive peak, N for a vertex-negative one — and numbered in sequence. The obligatory complex is P1, N1, P2 and, especially in children, a later N2. The peaks are also referred to by their typical adult latency: P1 (P50), N1 (N100), P2 (P200), N2 (N250).

These latencies dwarf those of the brainstem response — N1 alone arrives ten times later than the entire ABR — and so do the amplitudes. The cortical response is measured in microvolts rather than fractions of one, which is why it needs far fewer stimulus repetitions to record [19].

The components one by one

P1(~50–100 ms) is modest in adults but is the dominant, most robust peak in young children — so much so that it becomes the workhorse of paediatric work, where its latency tracks central maturation [14].

N1(~100 ms) is the largest deflection in the awake adult and the most studied cortical component, with multiple generators around auditory cortex [4]. It is exquisitely sensitive to stimulus onset, to the interval between stimuli, and to arousal — it fades in drowsiness and is unreliable in infants.

P2 (~180–200 ms) follows N1. In practice the N1–P2 amplitude — the peak-to-trough swing — is the single measure most often tracked, because it is large, repeatable and grows with stimulus level [5]. N2(~200–300 ms) is more prominent in children and adds little to routine adult testing.

What changes the waveform

The components are not fixed. As stimulus level rises, N1–P2 amplitude grows and latency shortens — the relationship that lets the response estimate threshold. Age reshapes the whole complex: the infant response is dominated by a broad positivity, N1 emerges only in later childhood, and latencies shorten as the cortex matures[15]. Faster stimulus rates and the listener drifting toward sleep both shrink the response, which is why timing and state are controlled so carefully during recording[6].

The acoustic change complex

The same onset machinery responds not only to a sound beginning but to a change withinan ongoing sound. A shift in frequency, intensity or spectrum partway through a continuous stimulus evokes a second N1–P2 — the acoustic change complex (ACC) [8]. It is elicited by both spectral and intensity changes [9].

The ACC is powerful because it is an objective index of discrimination, not just detection: the change response only appears if the auditory system registered the change. That makes it a natural tool for suprathreshold testing and for confirming that a hearing aid or implant delivers audible, distinguishable speech cues[10].

Electrodes and montage

Because the N1–P2 complex is largest at the top of the head, the active electrode sits at the vertex (Cz), referenced to the mastoids or linked earlobes, with a forehead ground. A single recording channel is enough for routine threshold work [18].

The response is measured in microvolts, so it tolerates a simpler montage and far fewer stimulus repetitions than the sub-microvolt ABR. The limiting factor is rarely the response — it is the background EEG noise sitting on top of it.

Stimuli: tones and speech tokens

Two stimulus families dominate. Tone bursts at the audiometric frequencies give frequency-specific threshold estimates, much as in behavioural audiometry. The stimulus needs a long enough rise time and a slow enough presentation rate to give a clean onset response; rate and rise time both shape the size of the recorded response [6].

Speech tokens — short natural sounds such as /m/, /g/ and /t/ — sample the low, mid and high frequency regions of speech. A cortical response to each token shows whether that part of the speech spectrum is audible, which is the basis of aided testing in infants [13]. Speech stimuli also bring the test closer to the everyday listening question that matters to patients [10].

Arousal state and free-field testing

The single biggest practical constraint is state. The late cortical response is generated in the cortex and weakens with drowsiness and sleep, so the patient must be awake, relaxed and quietly alert — reading or watching a silent, captioned video works well [5]. This is the mirror image of infant ABR, which is recorded during sleep.

For aided assessment the stimulus is delivered in the free field through a loudspeaker rather than an insert earphone, so the patient can be tested wearing their hearing aids or cochlear implant. The recording then reflects what the device actually delivers to the cortex.

Averaging and artefact

As with any evoked potential, the stimulus is repeated and the EEG epochs are averaged: the time-locked response adds up while random EEG averages toward zero. Because the cortical response is large, far fewer sweeps are needed than for the ABR. Epochs contaminated by blinks, movement or muscle activity are rejected, and the practical goal is to drive the leftover residual noise low enough to see the response clearly [11].

The principle

Cortical electric response audiometry (CERA) rests on a simple relationship known since the 1960s: as stimulus level falls, the cortical response shrinks and lengthens in latency, until at some level no repeatable response can be distinguished from the background noise [2]. The lowest level that still yields a clear, repeatable response estimates the threshold for that stimulus.

Run frequency by frequency with tone bursts, this builds an objective audiogram. In cooperative adults the cortical thresholds typically sit within about 5–10 dB of the behavioural thresholds — close enough to be clinically useful, with a small, predictable correction [6].

Why the cortical response, and not the ABR?

For threshold estimation in adults the late response has real advantages. It uses frequency-specific tonal stimuli without the rate and timing compromises of the brainstem response; it reflects processing all the way to the cortex; and it is the most direct objective analogue of the behavioural audiogram [5]. The auditory steady-state response (ASSR) is an alternative frequency-specific objective method — and it works in sleep — so the two are often considered together depending on the patient[7].

The trade-off is state: CERA needs an awake, cooperative patient, so it is most at home in older children and adults rather than in sleeping infants, where ABR and ASSR lead.

Non-organic hearing loss and medicolegal testing

The most established clinical niche is the objective confirmation of threshold when the behavioural audiogram is in doubt — suspected non-organic (functional) hearing loss, or compensation and medicolegal claims. Because the obligatory response cannot be voluntarily suppressed and gives frequency-specific thresholds in an awake adult, CERA is the reference method here[5]. A cortical threshold markedly better than the claimed behavioural threshold objectively exposes the exaggeration.

Aided cortical assessment

Delivered in the free field to speech tokens, the cortical response can confirm that amplified sound is audible at the cortex. With the patient wearing their hearing aid, a present response to /m/, /g/ and /t/ shows that the low-, mid- and high-frequency regions of speech are getting through[13]. This is especially valuable in infants, who cannot report what they hear: an objective check that a fitting is actually delivering audible speech [12].

Present, absent, or equivocal

A response is presentwhen a repeatable N1–P2 (or, in a child, P1) emerges clearly above the background, at the expected latency, and reproduces on a repeat average. It is absent when no such deflection rises out of the noise. The crucial third category is equivocal — a recording too noisy to call either way, which must be resolved by collecting more data, not by guessing.

Detection is fundamentally a signal-versus-noise decision. Because the response itself is large, the limiting variable is almost always the residual noise in the average. A flat trace from a noisy recording is not evidence of an absent response — it is evidence of a bad recording [11].

From visual reading to automated detection

Traditionally an experienced tester decided by eye whether a response was present — reliable in good hands, but subjective and hard to standardise. Automated statistical detection now supports that judgement: a multivariate statistic such as Hotelling’s T² tests whether the averaged waveform differs from flat noise across several time points, returning an objective present/absent decision with a known false-alarm rate[11].

Validated first in adults and then extended to infants, automated detection underpins clinical aided-CAEP systems and is especially valuable where the tester is not a cortical-response specialist[12]. It does not abolish the noise problem — a confident “absent” still requires the residual noise to have been driven low enough that a response would have shown if present.

What an absent response means — and what it doesn’t

An absent response at a given level means only that no cortical response was detected at that level under those conditions. It could mean the sound was inaudible — or that the patient drifted toward sleep, that the recording was too noisy, or that the stimulus was poorly chosen. State and noise must be excluded before an absent response is read as a threshold.

Conversely, a present response is strong evidence: sound reached and was processed by the cortex. This asymmetry is what makes the cortical response so useful in auditory neuropathy spectrum disorder, where the ABR is absent or grossly abnormal yet a present cortical response can still predict speech-perception potential [16]. The cortical response answers a question the brainstem response cannot[17].

P1 as a maturational biomarker

The infant cortical response looks nothing like the adult’s: it is dominated by a single broad positivity — an early form of P1 — while N1 emerges only later in childhood[15]. Crucially, P1 latency is not fixed. It is driven down by auditory experience as the cortex matures, falling steeply over the first years of life and then settling toward adult values.

Because that descent depends on stimulation reaching the cortex, P1 latency works as a biomarker of central auditory maturation. Plotted against age-appropriate norms, it tells us whether the central pathways are developing on schedule[14].

The sensitive period

The developing auditory cortex is maximally plastic during an early sensitive period. A child who receives auditory input early — through hearing aids or a cochlear implant — shows P1 latency catching up to normal; a child stimulated only after the window has largely closed tends to retain an abnormal, delayed P1, mirroring poorer spoken-language outcomes [14]. This is one of the physiological arguments for early identification and early implantation.

The same logic explains cross-modal reorganisation: when auditory cortex is deprived of input for too long, other senses can recruit it, and the cortical response reflects this altered development [15].

Using maturation clinically

In practice, P1 latency complements behavioural and device measures: it offers an objective, repeatable index of whether a fitting or implant is driving normal cortical development, and it can flag a child whose central maturation is lagging despite apparently adequate amplification. Combined with aided speech-token responses and the acoustic change complex, the cortical response gives a developmental picture no single behavioural test provides[10].