Sensory Thresholds

Abstract

Sensory thresholds mark the limits of perception: the least stimulus energy that reliably produces a sensory response, and the smallest change between two stimuli that can be told apart. The first is the absolute threshold, the second the difference threshold, and the systematic study of both founded psychophysics, the oldest quantitative branch of experimental psychology. Nineteenth-century work established two enduring regularities, that the just-noticeable difference grows in proportion to the standard stimulus, and that sensation magnitude is a lawful function of physical intensity, while the twentieth century reconceived the threshold itself as a decision under uncertainty rather than a fixed sensory barrier. Modern research replaces the static threshold with the psychometric function and estimates it efficiently with adaptive, Bayesian methods. Threshold measurement remains the working bridge between the physical stimulus and mental experience.

Keywords: sensory threshold, absolute threshold, difference threshold, Weber's law, psychometric function

A candle flame seen on a dark night at a distance of tens of kilometres; a single drop of perfume diffusing through a small house; the tick of a watch in a silent room a few metres away: the classic textbook illustrations of the absolute threshold dramatise a real and measurable fact, that sensation begins only once a stimulus crosses a lower limit. Below that limit the world produces no experience; just above it, experience appears, though not sharply. The measurement of these limits, and of the smallest detectable change in a stimulus, is the subject matter of psychophysics, the discipline that Gustav Fechner founded in 1860 to give the relation between body and mind a quantitative form. Its central instrument, the threshold, has been refined for over a century and remains the point at which a physical quantity is converted into a psychological one (Gescheider, 1988).

What Sensory Thresholds Are

MeSH defines sensory thresholds as “the minimum amount of stimulus energy necessary to elicit a sensory response.” The definition names the absolute threshold: the faintest light, quietest sound, or weakest concentration that a person can detect. A second, distinct limit is the difference threshold, the smallest change in a stimulus that can be told from the original, also called the just-noticeable difference. The two answer different questions, one about the onset of sensation and one about its resolution, but both are properties of the same perceptual system operating near its floor, which is why the National Library of Medicine indexes them together under the single heading.

Neither limit is a hard edge. Detection near the absolute threshold is probabilistic: a stimulus of a given intensity is caught on some trials and missed on others, and the “threshold” is defined by convention as the intensity detected on a fixed proportion of trials, usually one half or, in forced-choice designs, some criterion above chance. This graded character reflects the physics of the stimulus and the noisiness of the nervous system. The lower bound is remarkably low: under optimal conditions the dark-adapted human eye responds to only a handful of photons, and recent work using a single-photon source has shown that observers can report the arrival of a single photon at a rate slightly but reliably above chance, placing the visual absolute threshold at the physical limit set by one quantum of light (Tinsley et al., 2016).

Types of Sensory Thresholds

Beyond being a subject in its own right, Sensory Thresholds is a formal category in the National Library of Medicine's Medical Subject Headings, which places it at tree position F02.463.593.710, beneath Perception, and hangs its recognised narrower kinds beneath it. Most of these subtypes name the threshold of a particular modality or task; several are treated at length in their own articles. Table 1 lists the direct children of the descriptor.

Two cautions keep this taxonomy in its place. It is a classification for indexing, built to organise the literature, not a theory asserting that these subtypes are mutually exclusive or an exhaustive set of natural kinds; an auditory threshold and a difference threshold, for instance, describe the same ear measured two ways. And Signal Detection, Psychological sits here not as another modality but as the framework that dissolved the classical notion of a fixed threshold altogether, a reminder that MeSH organises topics as the literature indexes them rather than as a single coherent model of the mind.

Weber's Law and the Difference Threshold

The first quantitative regularity of the senses was found not at the absolute limit but at the difference threshold. Ernst Heinrich Weber observed in the 1830s that the just-noticeable difference between two weights, two lengths, or two tones is not a fixed amount but a fixed proportion of the standard against which it is judged. Lifting a 100-gram weight, a person needs about 2 grams added to notice a difference; lifting a 200-gram weight, about 4 grams. The ratio of the just-noticeable difference ΔI to the standard intensity I is approximately constant, a relation now written as Weber's law:

ΔI / I = k

The constant k, the Weber fraction, is a compact index of a modality's acuity: about 0.02 for judging brightness or lifted weight, higher for smell and taste, lower for the pitch of a tone. Weber's law holds well across the middle of the intensity range and breaks down near the absolute threshold, where a small additive constant must be added to I to keep the fraction stable (Gescheider, 1988). Its importance is less the precise fit than the discovery it embodies: that a mental event, the noticing of a difference, obeys a simple law expressed in physical units. The demonstration below lets the standard stimulus be varied so that the proportional growth of the just-noticeable difference can be read directly.

Fechner's Law, Stevens' Law, and Psychophysical Scaling

Gustav Fechner turned Weber's empirical fraction into a theory of the mind. Reasoning that each just-noticeable difference is a subjectively equal step, he integrated Weber's law to obtain a relation between the physical intensity of a stimulus and the magnitude of the sensation it evokes: sensation grows as the logarithm of intensity. Fechner's law, S = k log I, implies that equal ratios of stimulus intensity produce equal increments of sensation, so that a stimulus must be multiplied, not merely added to, for its felt magnitude to keep rising in equal steps. This compressive relation explains why a candle added to a dark room is dramatic while the same candle added to a bright room is imperceptible, and it dominated psychophysics for a century (Krueger, 1989).

Fechner's law rests on an assumption that S. S. Stevens contested: that all just-noticeable differences are subjectively equal. Using direct scaling methods, in which observers assign numbers to sensations or adjust one stimulus to be a stated multiple of another, Stevens argued that sensation magnitude grows not as the logarithm of intensity but as a power function, S = k I^a (Stevens, 1957). The exponent a is the signature of the modality: below one for compressive continua such as brightness and loudness, where sensation grows more slowly than intensity, and above one for expansive continua such as electric shock, where it grows faster. Stevens marshalled this evidence explicitly to displace Fechner (Stevens, 1961), and later work showed the exponent to be lawfully related to the dynamic range of the stimulus continuum (Teghtsoonian, 1971). The disagreement is not merely historical arithmetic: it turns on what counts as a valid scale of measurement, a question Stevens had already sharpened by classifying all measurement into nominal, ordinal, interval, and ratio scales (Stevens, 1946). The two laws have since been shown to be partly reconcilable, each capturing the data under different assumptions about how internal magnitudes are read out (Krueger, 1989). A modern critique goes further, arguing that the compound label Weber-Fechner law conflates Weber's well-attested empirical fraction with Fechner's contested logarithmic inference and should be retired as a misnomer (Algom, 2021). Table 2 sets the two laws side by side, and the demonstration that follows lets the exponent be varied to morph one curve into the other.

Measuring the Threshold: Psychophysical Methods

A threshold is not observed but estimated, and the history of psychophysics is in large part a history of methods for estimating it efficiently and without bias. Fechner bequeathed three classical procedures. In the method of limits the experimenter raises or lowers the stimulus in small steps until the observer's report changes, taking the crossover as the threshold; it is quick but prone to anticipation and habituation errors. In the method of constant stimuli a fixed set of intensities is presented many times each in random order, and the proportion of “yes” responses is plotted against intensity to yield the psychometric function, from which the threshold is read as the intensity detected on half the trials. In the method of adjustment the observer controls the stimulus directly and sets it to the point of just detection. The method of constant stimuli is the most accurate and the most laborious, because it spends many trials on intensities far from the threshold that contribute little to its estimate.

That inefficiency motivated adaptive procedures, which concentrate trials near the current threshold estimate by choosing each stimulus in light of the responses so far. The simplest is the staircase, formalised by Cornsweet, in which the stimulus steps down after a correct response and up after an error, so that the sequence oscillates around the threshold (Cornsweet, 1962). More powerful are the Bayesian adaptive methods: Watson and Pelli's QUEST maintains a probability distribution over the threshold and places each trial at its current best estimate, converging in far fewer trials (Watson & Pelli, 1983), and later procedures extended the idea to estimate the slope of the psychometric function as well as its location (Kontsevich & Tyler, 1999). Leek's review surveys the family and its trade-offs (Leek, 2001). Whatever the sampling method, the threshold and the reliability of its estimate depend on how the psychometric function is fitted; Wichmann and Hill gave the modern reference treatment, showing how the observer's occasional lapses must be modelled to avoid biasing the estimate and how bootstrap methods yield honest confidence intervals (Wichmann & Hill, 2001). Figure 1 shows the psychometric function and the quantities read from it.

Figure 1

The Psychometric Function and the Threshold Read From It

Is There a Sensory Threshold? Signal Detection

The classical threshold assumes a fixed sensory barrier: below it, nothing; above it, sensation. By the middle of the twentieth century this assumption was in trouble. Observers detect the same faint stimulus on some trials and miss it on others, and they report stimuli that were never presented; worse, how often they do either depends on their expectations and on the rewards and penalties attached to hits and false alarms, none of which should matter if the threshold were a fixed sensory fact. Tanner and Swets resolved the difficulty by importing statistical decision theory: the observer's task is not to cross a barrier but to decide, on each trial, whether a noisy internal signal is more likely to have arisen from noise alone or from signal-plus-noise (Tanner & Swets, 1954).

On this account, detection has two independent components. Sensitivity (indexed by d′) measures how far the signal-plus-noise distribution is separated from the noise distribution, a property of the observer and the stimulus. The criterion measures where the observer places the decision boundary, a strategic choice shaped by expectation and payoff. A cautious observer and a reckless one can have identical sensitivity yet very different hit and false-alarm rates, which is exactly what the classical threshold cannot represent. Swets pressed the point in a paper whose title asked directly whether there is a sensory threshold, and answered that the data are better described by a continuous decision process than by any fixed limit (Swets, 1961). Signal detection theory did not abolish the threshold as a practical measure, but it explained what a threshold estimate confounds, and it grounds the modern view that the psychometric function reflects a decision process whose form can be derived from the mechanisms of perceptual choice (Gold & Ding, 2013).

Worked Example

Consider a difference-threshold experiment in weight lifting. An observer is tested against a standard weight of 100 grams and reliably notices a change when 2 grams are added, so the just-noticeable difference is ΔI = 2 g and the Weber fraction is k = ΔI / I = 2 / 100 = 0.02. Weber's law predicts that this fraction stays constant as the standard changes. Against a 400-gram standard, then, the just-noticeable difference should be ΔI = k × I = 0.02 × 400 = 8 grams, four times the increment needed at 100 grams. The observer is no less acute; the same 2 percent resolution simply corresponds to a larger absolute step because the standard is larger.

The scaling laws make the parallel prediction for perceived magnitude. Under Fechner's law, S = k log I, each doubling of intensity adds the same increment of sensation: going from 100 to 200 grams and from 200 to 400 grams feel like equal steps, because the physical ratio is equal in both. Under Stevens' law with an exponent of a = 1.45 for lifted heaviness, doubling the weight multiplies the sensation by 2^1.45 ≈ 2.73, so heaviness grows slightly faster than weight. The two laws diverge most where they are easiest to test, at the extremes of intensity, which is why direct scaling and difference-threshold data pull toward different functional forms rather than settling the matter (Stevens, 1957; Krueger, 1989). The arithmetic underscores the conceptual point: a threshold is a ratio, not an absolute amount, and perceived magnitude is a lawful but non-linear transform of the physical scale.

Discussion

The sensory threshold has proven one of the most durable constructs in experimental psychology, not because it names a fixed thing but because measuring it has repeatedly forced the field to sharpen its concepts. Weber's constant fraction and Fechner's logarithm gave psychology its first quantitative laws and, with them, the founding claim that mental events can be measured through their physical correlates (Gescheider, 1988). Stevens' power law and his analysis of measurement scales showed that the choice of scaling method is not a technical detail but a theoretical commitment about what a number attached to a sensation can mean (Stevens, 1946; Stevens, 1957). Signal detection theory then revealed that the classical threshold silently confounds sensitivity with decision, and replaced the barrier with a continuous process whose two components can be measured apart (Tanner & Swets, 1954; Swets, 1961).

The unifying object across these shifts is the psychometric function. It absorbs the classical threshold as one of its parameters, expresses signal detection's continuity in its smooth rise, and makes the estimation problem statistical rather than metaphysical: the question is no longer “where is the threshold?” but “what function best describes how detection depends on intensity, and how precisely can it be estimated?” (Wichmann & Hill, 2001). That reframing links psychophysics to the study of perceptual decision-making, where the same function is derived from the dynamics of evidence accumulation (Gold & Ding, 2013). What began as a search for the point where sensation switches on has become the measurement of a graded, decision-laden mapping between stimulus and experience.

Current Directions

Contemporary threshold research is largely methodological and computational, aimed at extracting more information from fewer trials. Watson's QUEST+ generalises the Bayesian adaptive approach to estimate several parameters of the psychometric function at once, and to handle experiments with multiple stimulus dimensions and response categories, so that slope, threshold, and lapse rate can be recovered together within a single efficient design (Watson, 2017). In parallel, advances in fitting have made estimation more robust: hierarchical and Bayesian treatments accommodate the overdispersion of real data, in which observers are more variable than a binomial model assumes, and return credible intervals rather than point estimates alone (Schütt et al., 2016). Open, reproducible tools have followed, packaging these fitting methods so that any laboratory can apply them consistently (Linares & López-Moliner, 2016).

At the other extreme, the absolute threshold has been pushed to its physical floor. The demonstration that humans can detect single photons closes a question open since the 1940s and turns the visual system into a testbed for quantum-limited detection, where the psychophysics of the threshold meets the physics of the stimulus (Tinsley et al., 2016). Between these poles, the continuing project is to connect the descriptive psychometric function to the mechanisms that generate it, so that a threshold is read not as an isolated number but as a summary of how a particular perceptual system decides under uncertainty (Gold & Ding, 2013).

Glossary

Absolute threshold.

The least stimulus energy that can be detected, defined by convention as the intensity detected on a fixed proportion of trials, usually one half.

Adaptive procedure.

A method that chooses each stimulus in light of previous responses, concentrating trials near the current threshold estimate for greater efficiency.

Criterion.

In signal detection theory, the internal decision boundary an observer adopts; a strategic setting shaped by expectation and payoff, independent of sensitivity.

Difference threshold.

The smallest change in a stimulus that can be discriminated from the original; the just-noticeable difference.

Fechner's law.

The proposal that sensation magnitude grows as the logarithm of stimulus intensity, derived by integrating Weber's law under the equal-JND assumption.

Just-noticeable difference.

The increment in a stimulus that is detected as a change on a criterion proportion of trials; the operational form of the difference threshold, abbreviated JND.

Method of constant stimuli.

A classical procedure presenting a fixed set of intensities many times in random order to trace the full psychometric function; accurate but laborious.

Method of limits.

A classical procedure that raises or lowers the stimulus in steps until the observer's report changes, taking the crossover as the threshold.

Psychometric function.

The curve relating the proportion of detections or correct responses to stimulus intensity; its parameters include the threshold, the slope, and the lapse rate.

Psychophysics.

The quantitative study of the relation between physical stimuli and the sensations and perceptions they produce; founded by Fechner in 1860.

Sensitivity.

In signal detection theory, the separation between the noise and signal-plus-noise distributions, indexed by d′; a property of observer and stimulus, independent of the criterion.

Staircase method.

An adaptive procedure that steps the stimulus down after a correct response and up after an error, so the sequence oscillates around the threshold.

Stevens' power law.

The proposal that sensation magnitude grows as a power function of stimulus intensity, with an exponent that varies by modality and can be below or above one.

Weber fraction.

The constant ratio of the just-noticeable difference to the standard stimulus intensity; a compact index of a modality's discriminative acuity.

Weber's law.

The regularity that the just-noticeable difference is a constant fraction of the standard stimulus; the first quantitative law of the senses.

Key Researchers

Gustav Theodor Fechner (1801-1887). Physicist and philosopher at the University of Leipzig; founded psychophysics in Elemente der Psychophysik (1860), formalising the threshold and proposing the logarithmic law relating sensation to stimulus intensity. Wikipedia

David Marvin Green (1932-2022). Auditory psychophysicist; with John Swets wrote Signal Detection Theory and Psychophysics (1966), the canonical text that made sensitivity and response bias standard tools across hearing and perception. National Academy of Sciences

Denis G. Pelli (contemporary). Vision scientist at New York University; with Andrew Watson devised QUEST, the Bayesian adaptive method that places each trial near the current threshold estimate and remains a standard psychophysical procedure. Faculty Page - Wikipedia - ORCID

Stanley Smith Stevens (1906-1973). Psychophysicist at Harvard University; proposed the power law of sensation, challenging Fechner, and defined the nominal, ordinal, interval, and ratio taxonomy of measurement scales. Faculty Page - Wikipedia

John A. Swets (1928-2016). Mathematical psychologist; with Wilson Tanner brought statistical decision theory to detection, showing that the classical threshold is better modelled as a decision separating sensitivity from criterion. Wikipedia

Andrew B. Watson (contemporary). Vision scientist, Chief Vision Scientist at Apple and formerly at NASA Ames; devised QUEST with Pelli and the QUEST+ multidimensional extension, and founded the Journal of Vision. Optica Biography - Google Scholar

Ernst Heinrich Weber (1795-1878). Anatomist and physiologist at the University of Leipzig; discovered that the just-noticeable difference is a constant fraction of the standard stimulus, the empirical law on which Fechner built psychophysics. Wikipedia

Felix A. Wichmann (contemporary). Vision scientist at the University of Tübingen; gave the modern reference treatment of fitting the psychometric function, including the role of the lapse rate and bootstrap confidence intervals. Faculty Page - Google Scholar - ORCID

Frequently Asked Questions

What is a sensory threshold?
A sensory threshold is a limit of perception. The absolute threshold is the least stimulus energy that can be detected, and the difference threshold is the smallest change between two stimuli that can be discriminated. Both are properties of a perceptual system operating near its floor, which is why MeSH indexes them together (Gescheider, 1988).

What is the difference between the absolute and difference thresholds?
The absolute threshold concerns the onset of sensation, the faintest detectable stimulus. The difference threshold concerns resolution, the smallest change that can be told from the original, also called the just-noticeable difference. One asks whether anything is there; the other asks whether two things differ.

What is Weber's law?
Weber's law states that the just-noticeable difference between two stimuli is a constant fraction of the standard stimulus, so that a heavier weight or a brighter light requires a proportionally larger change to be noticed. The ratio, the Weber fraction, indexes a modality's acuity (Gescheider, 1988).

How do Fechner's law and Stevens' law differ?
Fechner's law holds that sensation grows as the logarithm of intensity, derived indirectly from Weber's law. Stevens' law holds that it grows as a power function, measured directly by magnitude estimation, with an exponent that varies by modality. The two capture the data under different assumptions about how internal magnitudes are read out (Stevens, 1957; Krueger, 1989).

How is a threshold actually measured?
By presenting stimuli of varying intensity and tracing the psychometric function, the curve relating detection to intensity. Classical methods include limits, constant stimuli, and adjustment; modern experiments use adaptive procedures such as the staircase and Bayesian methods like QUEST that concentrate trials near the threshold (Cornsweet, 1962; Watson & Pelli, 1983).

Is there really a fixed sensory threshold?
Not in the classical sense. Signal detection theory showed that detecting a faint stimulus is a decision under uncertainty, in which sensitivity and the observer's response criterion are separate. The apparent threshold confounds the two, so it is better treated as a convenient summary than as a fixed sensory barrier (Tanner & Swets, 1954; Swets, 1961).

What is the lowest stimulus a human can detect?
For vision it is essentially a single photon. Under dark adaptation the eye responds to a handful of quanta, and an experiment using a single-photon source showed observers reporting the arrival of one photon at a rate reliably above chance, placing the visual absolute threshold at the physical limit of light (Tinsley et al., 2016).

Why does the psychometric function matter more than the threshold point?
Because it contains the threshold as just one parameter alongside the slope and lapse rate, and it expresses the graded, probabilistic nature of detection that a single point cannot. Fitting it well, and estimating its uncertainty, is the central technical problem of modern threshold measurement (Wichmann & Hill, 2001).

References