Auditory Threshold

Abstract

The auditory threshold is a type of auditory perception: the faintest sound a listener can reliably detect, and the quantity on which the measurement of hearing rests. It is not a single number but a family of measures — the minimum audible pressure at the eardrum, the minimum audible field in free space, the level relative to a normal-hearing reference, and the raised threshold produced by a competing masker. Plotted against frequency it traces the audibility curve, most sensitive near 2–4 kHz and rising steeply toward low and high frequencies. Signal-detection analysis showed the threshold is a statistical boundary set by the listener's criterion as much as by sensory sensitivity, so modern procedures estimate it with adaptive staircases that target a fixed point on the psychometric function rather than a fixed sensation.

Keywords: auditory threshold, absolute threshold, audibility curve, adaptive staircase, masking

What the Auditory Threshold Is

The auditory threshold is the smallest sound intensity that evokes an auditory sensation — operationally, the level a listener detects on some agreed proportion of presentations. It is the auditory special case of the sensory threshold, the general psychophysical boundary between a stimulus too weak to perceive and one strong enough to report, and it is measured within the wider practice of psychophysics and its clinical offshoot, audiometry. Because sound energy spans an enormous range, the threshold is expressed logarithmically, in decibels, against a fixed reference pressure of 20 micropascals, the quantity defined as 0 dB sound pressure level (SPL).

Two things make the auditory threshold more than a single reading. First, it depends strongly on the frequency of the tone: the ear is far more sensitive to a 3 kHz tone than to a 100 Hz or a 15 kHz tone, so the threshold is properly a curve across the audible range rather than a point. Second, it depends on how the sound reaches the ear and on what else is present: the faintest detectable pressure measured at the eardrum through an earphone differs from the faintest detectable level of a sound played in open space, and both rise sharply when a competing sound masks the target. The measures below name these distinctions precisely.

Underlying all of them is a deeper problem the early tradition did not anticipate. A threshold implies a sharp cliff — silence below, sensation above — but detection is in fact probabilistic, climbing gradually from chance to certainty as level rises. Where on that gradual psychometric function one draws the “threshold” is a convention, and, as signal-detection theory later showed, it reflects the listener's willingness to say yes as much as the sensitivity of the ear. Table 1 sets out the principal measures and the quantity each reports.

Table 1

Measures of the Auditory Threshold

Note. The threshold is a family of related quantities, not one number; which is meant depends on how the sound is delivered and against what reference it is scored. Full citations appear in the References (Sivian & White, 1933; Fletcher & Munson, 1933; Carhart & Jerger, 1959; Suzuki & Takeshima, 2004).

Types of Auditory Threshold

MeSH classifies Auditory Threshold under two parents at once — auditory perception in the perceptual tree and sensory thresholds in the general psychophysics tree — which is an indexing choice, not a claim that the threshold belongs to two different kinds. Beneath the descriptor it carries a single narrower term:

- Auditory Fatigue — a temporary rise in the auditory threshold following exposure to sound, so that a tone detectable before exposure is no longer detectable until the ear recovers. It is the threshold-shift phenomenon that connects the measurement of hearing to noise damage, and the measure that makes a temporary threshold shift visible.

These subtypes partition the descriptor by cause of threshold change rather than by a property of the listener, so they are not mutually exclusive with the measures in Table 1: auditory fatigue is a shift in whichever threshold — MAP, MAF, or HL — is being tracked. The classification is a bibliographic convenience for indexing the hearing literature, not a theory of the threshold itself.

The Audibility Curve

When the absolute threshold is measured across the audible range and plotted against frequency, it traces a shallow trough: the audibility curve, also called the threshold of hearing. Sensitivity is greatest between roughly 2 and 5 kHz, where a young healthy ear can detect pressures at or even below the 20 micropascal reference, and falls off steeply on either side. At 100 Hz the threshold is some 40 dB higher than at 3 kHz; above about 15 kHz it rises almost vertically toward the upper edge of hearing. The resonance of the ear canal, which boosts pressure at the eardrum near 3 kHz, and the mechanics of the middle ear together shape this trough.

Sivian and White measured the two canonical versions of the curve and exposed a puzzle between them. The minimum audible field (MAF), the faintest free-field level detectable with both ears, lies several decibels below the minimum audible pressure (MAP) measured at the eardrum through an earphone — the “missing 6 dB” — a discrepancy later traced to binaural summation, ear-canal and head diffraction in the free field, and physiological noise heard under an earphone (Sivian & White, 1933). The audibility curve is also the lowest member of a family: the equal-loudness contours map the levels that sound equally loud across frequency, and the threshold curve is their 0-phon floor. Fletcher and Munson measured the first such contours and defined loudness level in phons (Fletcher & Munson, 1933); Suzuki and Takeshima's modern re-measurement became the basis of the international standard contours (Suzuki & Takeshima, 2004).

Measuring the Threshold

The oldest method is the method of limits: the experimenter presents a tone in ascending or descending series, stepping the level until the listener's report changes from inaudible to audible or back, and averages the crossing points. It is quick but biased — listeners anticipate the change in an expected direction — and it wastes most trials far from the threshold. Two developments made the measurement both self-recording and efficient.

The first was tracking audiometry. In Bekesy's tracking audiometer the listener holds a button to attenuate a tone while it is audible and releases it when the tone disappears; the level therefore oscillates automatically around the threshold, which is read as the midline of the oscillation, while the frequency sweeps slowly across the range to trace the whole audibility curve without an experimenter in the loop. The second was the clinical standardisation of the method of limits into a fixed bracketing rule. Carhart and Jerger compared procedures and recommended the Hughson–Westlake “down 10, up 5” staircase: after each detected tone the level drops 10 dB, after each miss it rises 5 dB, and the threshold is the lowest level detected on at least half of ascending presentations (Carhart & Jerger, 1959). This remains the method of most pure-tone audiometry.

The most efficient laboratory method is the adaptive staircase. Levitt generalised the up–down rule so that the step direction depends on a short run of responses rather than a single one: in a two-down, one-up rule the level falls only after two consecutive correct detections and rises after any miss, which drives the track toward the level detected 70.7% of the time rather than the 50% point of a simple up–down run (Levitt, 1971). The threshold is estimated as the mean of the level at the reversals, the points where the track changes direction, after the first few are discarded. Different target percentages are reached by choosing the run length, which is why transformed up–down rules became the standard tool of psychoacoustics.

Is There a True Threshold?

Classical psychophysics treated the threshold as a fixed sensory boundary blurred only by inattention and momentary noise. Signal-detection theory dissolved that picture. Swets, Tanner, and Birdsall argued that a faint tone is detected against a background of internal noise, so that on any trial the listener compares a noisy sensory sample against an internal criterion and says yes when it exceeds that criterion (Swets, Tanner, & Birdsall, 1961). Detection performance then has two independent components: sensitivity (how far the signal shifts the sensory sample, measured by d′) and bias (where the criterion sits, how readily the listener says yes). A cautious listener and a liberal one can have identical ears yet report very different “thresholds,” because the apparent threshold moves with the criterion.

This reframing, developed in full on the signal-detection theory page, is why a bare yes–no threshold is unreliable and why modern procedures use forced choice. If the tone is presented in one of two clearly marked intervals and the listener must choose which contained it, guessing is distributed evenly and criterion bias is largely removed, so the measured level reflects sensitivity alone. The adaptive staircase of the previous section is almost always run this way. The “threshold” that results is explicitly a chosen point on the psychometric function — the 70.7%-correct level in a two-down, one-up rule — not a sensory cliff.

Masking and the Masked Threshold

A threshold measured in quiet is only the simplest case. When a second sound, the masker, is present, the threshold for the target rises, sometimes by tens of decibels; the elevated level is the masked threshold, and the difference from the quiet threshold is the amount of masking. Masking is frequency-specific in a revealing way: a masker raises the threshold most for targets close to its own frequency and less for targets far away, and the region over which it is effective widens as the masker level increases. Fletcher's work at Bell Labs made this the basis of the critical band — the finding that only masker energy within a limited band around the target frequency contributes to masking it (Fletcher & Munson, 1933).

The critical band is the signature of the ear's action as a bank of overlapping filters: each place on the basilar membrane responds to a narrow range of frequencies, so a target is masked only to the extent that the masker drives the same filter. Masking patterns are also asymmetric — a low-frequency masker spreads its effect upward in frequency far more than downward, the “upward spread of masking” — because of the shape of the travelling wave in the cochlea. The masked threshold is thus not an artefact to be avoided but a probe of auditory frequency analysis, and it is central to how audio compression decides which sounds can be discarded unheard.

Hearing Level and the Clinical Audiogram

A threshold in dB SPL mixes two things a clinician wants to separate: the listener's hearing and the frequency-dependence of the ear itself. Because the normal threshold is some 40 dB higher at 100 Hz than at 3 kHz, a raw SPL audiogram of a normal ear is a deep curve, and a real hearing loss is hard to read against it. Hearing level (dB HL) removes the curve by referencing each frequency to the normal-hearing median: 0 dB HL is normal threshold at every frequency, so a normal audiogram plots as a flat line near the top and any downward deflection is loss in decibels. The reference values that define 0 dB HL were fixed by standards bodies from population measurements and refined by work like Carhart and Jerger's on the clinical method (Carhart & Jerger, 1959).

The audiogram built from these thresholds is the primary clinical record of hearing, read for the degree of loss (how far thresholds fall below 0 dB HL), its configuration across frequency (sloping, flat, notched), and its type (conductive, sensorineural, or mixed, separated by comparing air-conducted and bone-conducted thresholds). The pure-tone threshold, measured one frequency at a time by the staircase rule, is the single number from which this entire diagnostic picture is assembled.

Worked Example

Consider a two-down, one-up adaptive staircase estimating a listener's threshold for a 1 kHz tone, run as a two-interval forced choice so that bias is removed. The rule drives the track toward the level the listener detects 70.7% of the time, because the level falls only after two correct trials in a row — a descent requires a joint probability of 0.5, and the square root of 0.5 is 0.707.

The track changes direction at a series of reversals. Suppose, after the run has settled, the reversal levels (in dB HL) are:

20, 10, 18, 12, 17, 13, 16, 14

The first two reversals are discarded as part of the initial homing-in, and the threshold is taken as the mean of the last six:

threshold = (18 + 12 + 17 + 13 + 16 + 14) / 6 = 90 / 6 = 15 dB HL

The estimate is stable: averaging only the last four reversals (17, 13, 16, 14) gives (17 + 13 + 16 + 14) / 4 = 60 / 4 = 15 dB HL, the same value, because the track is oscillating symmetrically about 15 dB. That convergence is the point of the method: the staircase spends almost all its trials near the threshold, and the reversal mean reads it off directly.

To see what the decibel value means in pressure, recall that 0 dB SPL is the reference pressure of 20 micropascals. A threshold measured at 20 dB SPL corresponds to a pressure of 20 × 10^(20/20) = 200 micropascals — ten times the reference pressure, since every 20 dB is a factor of ten in pressure.

Current Directions

The clinical audiogram stops at 8 kHz, but hearing extends to around 20 kHz, and the thresholds in that extended high-frequency range are proving to carry information the standard audiogram misses. Mishra and colleagues showed that listeners with a normal standard audiogram can nonetheless have elevated extended high-frequency thresholds, and that those thresholds predict difficulty understanding speech in noise — evidence that a “normal” threshold up to 8 kHz can hide a real deficit (Mishra, Saxena, & Rodrigo, 2022). A recent review draws the same conclusion for ageing listeners, where extended high-frequency thresholds rise earliest and may give the earliest measurable sign of age-related loss (Gottfriedová et al., 2024).

A parallel strand concerns thresholds that stay normal while hearing is nonetheless impaired. In cochlear synaptopathy, or “hidden hearing loss,” the synapses between hair cells and the auditory nerve are damaged while the hair cells that set the quiet threshold survive, so the audiometric threshold is normal yet supra-threshold coding — the perception of speech in noise — is degraded. Both lines of work share a lesson that signal-detection theory first taught: the detection threshold measured in quiet, for all its clinical centrality, captures only part of what the auditory system does, and the frontier is in the measures that go beyond it.

Discussion

The auditory threshold began as the simplest possible question — what is the faintest sound one can hear — and turned out to organise a century of hearing science. The answer is never a single number: it is a curve across frequency, it depends on how the sound is delivered and on what competes with it, and, most consequentially, it is a statistical boundary rather than a physical one. The arc from the method of limits through Bekesy's self-tracking audiometer to Levitt's adaptive staircase is a steady refinement of how to locate a point on a probabilistic function efficiently and without bias, and signal-detection theory is the explanation of why that point was so elusive in the first place.

What the threshold buys, in return for this care, is extraordinary: a single level, measured one frequency at a time, from which the clinical audiogram, the diagnosis of loss type, and the calibration of every audio device are built. The current frontier — extended high frequencies and hidden hearing loss — is a reminder that the quiet detection threshold, though foundational, is a floor and not a ceiling: there is hearing it does not measure, and the measures that reach it are where the field is now moving.

Common Misconceptions

The auditory threshold is a single fixed level.

It is a curve across frequency and a family of measures. The ear is roughly 40 dB more sensitive near 3 kHz than at 100 Hz, and the threshold differs depending on whether it is measured at the eardrum (MAP) or in a free field (MAF), and on what else is sounding (Sivian & White, 1933).

Below the threshold there is silence, above it there is sound.

Detection climbs gradually from chance to certainty as level rises. The “threshold” is a chosen point on that psychometric function — conventionally the 70.7%-correct level in a two-down, one-up rule — and signal-detection theory shows the apparent level also depends on the listener's criterion, not sensitivity alone (Swets, Tanner, & Birdsall, 1961).

0 dB is the quietest possible sound, or silence.

0 dB SPL is simply a reference pressure of 20 micropascals, chosen near the normal threshold at 1 kHz; a healthy ear can detect pressures below it near 3 kHz, giving negative dB SPL thresholds, as the standardised free-field threshold of hearing shows a trough below 0 dB SPL between 3 and 4 kHz (Suzuki & Takeshima, 2004). And 0 dB HL means “normal threshold at this frequency,” not no sound at all.

A normal audiogram means hearing is intact.

The standard audiogram measures quiet detection only to 8 kHz. Listeners with a normal audiogram can have raised extended high-frequency thresholds, or cochlear synaptopathy, and struggle to understand speech in noise — hearing the threshold does not capture (Mishra, Saxena, & Rodrigo, 2022).

Glossary

Absolute threshold.

The smallest stimulus intensity a listener can detect on an agreed proportion of presentations; for hearing, the faintest audible sound at a given frequency.

Adaptive staircase.

A threshold procedure in which the level of each trial depends on previous responses, stepping down after correct detections and up after misses so that most trials fall near the threshold.

Audibility curve.

The absolute threshold of hearing plotted against frequency; a shallow trough most sensitive near 2–5 kHz and the 0-phon floor of the equal-loudness contours. Also called the threshold of hearing.

Auditory fatigue.

A temporary rise in the auditory threshold following sound exposure, so that a previously detectable tone cannot be heard until the ear recovers; the basis of the temporary threshold shift. The one MeSH subtype of auditory threshold.

Criterion.

In signal-detection theory, the internal level a noisy sensory sample must exceed before the listener reports a signal; its placement (bias) shifts the apparent threshold independently of sensitivity.

Critical band.

The limited band of frequencies around a target within which a masker's energy contributes to masking it; the signature of the ear's action as a bank of overlapping filters.

Decibel sound pressure level (dB SPL).

A logarithmic measure of sound pressure against a fixed reference of 20 micropascals, defined as 0 dB SPL; every 20 dB is a tenfold change in pressure.

Equal-loudness contour.

A curve joining the levels across frequency that sound equally loud; the contours are labelled in phons, and the audibility curve is their lowest, 0-phon member.

Hearing level (dB HL).

A threshold expressed relative to the normal-hearing median at each frequency, so that 0 dB HL is normal threshold everywhere and a normal audiogram plots as a flat line.

Hughson–Westlake procedure.

The standard clinical staircase for pure-tone thresholds, dropping 10 dB after a detected tone and raising 5 dB after a miss; standardised in the form recommended by Carhart and Jerger.

Masked threshold.

The elevated threshold for a target measured in the presence of a competing masker; the difference from the quiet threshold is the amount of masking.

Method of limits.

A classical procedure that presents a stimulus in ascending or descending series and records the level at which the report changes between detectable and undetectable; prone to anticipation bias.

Minimum audible field (MAF).

The faintest free-field sound level detectable with both ears, the listener facing the source; lies several decibels below the minimum audible pressure.

Minimum audible pressure (MAP).

The faintest sound pressure detectable at the eardrum, delivered monaurally through an earphone; the earphone-based form of the absolute threshold.

Phon.

The unit of loudness level: a tone's loudness level in phons is the dB SPL of an equally loud 1 kHz tone, so the equal-loudness contours are labelled by phon value.

Reversal.

A point in an adaptive staircase where the track changes direction; the threshold is estimated as the mean level of the reversals after the first few are discarded.

Sensitivity (d′).

In signal-detection theory, the separation between the internal distributions for signal-plus-noise and noise alone; a measure of detection ability independent of the listener's criterion.

Tracking audiometry.

A self-recording method in which the listener attenuates a tone while it is audible and releases it when it disappears, so the level oscillates around the threshold as frequency sweeps; the Bekesy audiometer.

Transformed up–down rule.

An adaptive rule in which the step direction depends on a run of responses rather than one; a two-down, one-up rule targets the 70.7%-correct level, other run lengths other percentages.

Key Researchers

Georg von Bekesy

(1899-1972). Won the 1961 Nobel Prize in Physiology or Medicine for showing how the travelling wave along the basilar membrane maps each sound frequency to a place in the cochlea, the mechanical basis of why the threshold depends so sharply on frequency. His tracking audiometer, in which the listener keeps a swept tone near their own threshold, remains the archetype of a self-recording threshold method. See his Wikipedia biography and Wikidata record.

Raymond Carhart

(1912-1975). Regarded as the father of audiology, he standardised the clinical measurement of the pure-tone threshold. The Carhart–Jerger refinement of the Hughson–Westlake bracketing procedure is still the method of most pure-tone audiometry, and the “Carhart notch,” a threshold dip near 2 kHz, bears his name. See his Wikipedia biography and Wikidata record.

Harvey Fletcher

(1884-1981). Founded quantitative psychoacoustics at Bell Labs. With Munson he measured the first equal-loudness contours and defined loudness level in phons, fixing the frequency weighting of the audibility curve, and his work on masking established the critical band that governs the masked threshold. See his Wikipedia biography and Wikidata record.

Brian C. J. Moore

(b. 1946). Emeritus Professor of Auditory Perception at the University of Cambridge and author of the standard textbook An Introduction to the Psychology of Hearing. His work on frequency selectivity, the auditory filter, and masking underpins how absolute and masked thresholds are modelled and measured. See his faculty page, Google Scholar profile, and Wikipedia biography.

Andrew J. Oxenham

(living). Professor of Psychology at the University of Minnesota, where he leads the Auditory Perception and Cognition Laboratory. His research on pitch, masking, and cochlear-implant hearing addresses the processes that set masked and frequency-specific thresholds. See his faculty page, Google Scholar profile, and Wikidata record.

Christopher J. Plack

(living). Ellis Llwyd Jones Professor of Audiology at the University of Manchester and author of The Sense of Hearing. His work on the physiological basis of the threshold includes cochlear synaptopathy, or hidden hearing loss, in which supra-threshold coding degrades while the standard audiometric threshold stays normal. ORCID 0000-0002-2987-5332; faculty page; Google Scholar profile.

Frequently Asked Questions

What is the auditory threshold?

The auditory threshold is the faintest sound a listener can reliably detect. Because the ear's sensitivity varies with frequency, it is properly a curve across the audible range rather than a single level, and it depends on how the sound is delivered and on what else is sounding. It is expressed in decibels against a reference pressure of 20 micropascals (0 dB SPL).

Why is the threshold lowest around 3 kHz?

The ear canal resonates near 3 kHz, boosting the pressure that reaches the eardrum, and the mechanics of the middle ear transmit mid frequencies most efficiently. Together these make the ear most sensitive between about 2 and 5 kHz, which is also the range carrying most of the information in speech.

What is the difference between dB SPL and dB HL?

dB SPL measures physical sound pressure against a fixed reference, so a normal threshold plotted in dB SPL traces the deep audibility curve. dB HL references each frequency to the normal-hearing median, so 0 dB HL is normal threshold everywhere and a normal audiogram plots as a flat line, making a hearing loss easy to read as a downward deflection.

How is the threshold actually measured?

Modern methods use an adaptive staircase: the sound level steps down after correct detections and up after misses, so most trials land near the threshold, which is read off as the mean level at the direction-changing reversals. Clinically, the Hughson–Westlake “down 10, up 5” bracketing rule is used; in the laboratory, transformed up–down rules target a chosen point on the psychometric function.

Is the threshold a sharp cliff between silence and sound?

No. Detection rises gradually from chance to certainty as level increases, so the “threshold” is a conventional point on that psychometric function. Signal-detection theory further shows that where the point appears depends on the listener's criterion — how readily they say yes — not on sensitivity alone, which is why forced-choice procedures are preferred.

What is a masked threshold?

It is the threshold for a target measured while a competing sound, the masker, is present. The masker raises the threshold most for targets near its own frequency, defining the critical band, because the ear analyses sound as a bank of overlapping frequency filters. The amount a masker elevates the threshold is the measure of masking.

Can hearing be impaired with a normal threshold?

Yes. The standard audiogram measures quiet detection only to 8 kHz, and a listener can have a normal audiogram yet raised extended high-frequency thresholds or cochlear synaptopathy (“hidden hearing loss”), and struggle to understand speech in noise. The quiet detection threshold is foundational but does not capture everything the auditory system does.

What is auditory fatigue?

Auditory fatigue is a temporary rise in the threshold after sound exposure, so that a previously audible tone cannot be heard until the ear recovers — a temporary threshold shift. It is the single narrower term MeSH files under auditory threshold, and it links the measurement of hearing to noise-induced damage.

References

Carhart, R., & Jerger, J. F. (1959). Preferred method for clinical determination of pure-tone thresholds. Journal of Speech and Hearing Disorders, 24(4), 330-345. https://doi.org/10.1044/jshd.2404.330

Fletcher, H., & Munson, W. A. (1933). Loudness, its definition, measurement and calculation. The Journal of the Acoustical Society of America, 5(2), 82-108. https://doi.org/10.1121/1.1915637

Gottfriedová, N., Kovalová, M., Škerková, M., & Mrázková, E. (2024). Extended high-frequency audiometry in the elderly: A narrative review. The Journal of International Advanced Otology, 20(4), 358-364. https://doi.org/10.5152/iao.2024.231217

Levitt, H. (1971). Transformed up-down methods in psychoacoustics. The Journal of the Acoustical Society of America, 49(2B), 467-477. https://doi.org/10.1121/1.1912375

Mishra, S. K., Saxena, U., & Rodrigo, H. (2022). Extended high-frequency hearing impairment despite a normal audiogram. Ear and Hearing, 43(3), 822-835. https://doi.org/10.1097/AUD.0000000000001140

Sivian, L. J., & White, S. D. (1933). On minimum audible sound fields. The Journal of the Acoustical Society of America, 4(4), 288-321. https://doi.org/10.1121/1.1915608

Suzuki, Y., & Takeshima, H. (2004). Equal-loudness-level contours for pure tones. The Journal of the Acoustical Society of America, 116(2), 918-933. https://doi.org/10.1121/1.1763601

Swets, J. A., Tanner, W. P., & Birdsall, T. G. (1961). Decision processes in perception. Psychological Review, 68(5), 301-340. https://doi.org/10.1037/h0040547