Psychophysics

Abstract

Psychophysics, which MeSH classifies within the behavioral sciences, is the quantitative study of the relationship between physical stimuli and the sensations and perceptions they evoke. Founded by Gustav Fechner in 1860, it supplies the measurement backbone of perception research: the absolute and difference thresholds, the lawful scaling of sensation magnitude, and rigorous procedures for estimating both. Weber's law fixes the difference threshold as a constant fraction of intensity; Fechner's logarithmic law and Stevens' power law compete to describe how sensation grows; signal detection theory separates sensitivity from decisional bias. Modern work fits psychometric functions with Bayesian rigor and grounds the classical laws in neural mechanism, making psychophysics as central to contemporary sensory science as it was to its nineteenth-century origins.

Keywords: psychophysics, Weber's law, threshold, power law, signal detection

The discipline begins with a deceptively simple question: how does the measurable world outside the observer map onto the graded world of experience inside (Fechner, 1860/1966)? Psychophysics answers it not with introspection but with controlled trials, forced-choice responses, and the mathematics of the sensory thresholds that separate the detectable from the imperceptible. Its methods and laws underwrite nearly every downstream claim in the science of perception.

What Psychophysics Is

Psychophysics is the branch of experimental psychology that relates physical stimulus dimensions (luminance, sound pressure, mass, concentration) to the psychological dimensions of sensation and perception they produce. Gustav Fechner coined the term and gave the field its program in the Elemente der Psychophysik of 1860, proposing that the relation between the outer and inner worlds could be made exact through measurement rather than speculation (Fechner, 1860/1966). His insight was methodological as much as theoretical: because sensation cannot be measured on a ruler, it must be inferred from the observer's discriminations, and those discriminations can be counted.

Two foundational quantities organize the field. The absolute threshold is the smallest stimulus intensity an observer can reliably detect against a null background. The difference threshold, or difference threshold, is the smallest change in intensity that can be told apart from a standard, a quantity Fechner named the just-noticeable difference. Both are statistical rather than fixed: detection probability rises gradually with intensity, so a threshold is defined as the intensity yielding a criterion level of performance, conventionally the midpoint of that rising curve (Gescheider, 1997).

Figure 1

From Stimulus to Sensation: The Psychophysical Chain

Types of Psychophysics

MeSH places Psychophysics (D011601) over two narrower descriptors, shown in Table 1. They partition the field by sensory domain and by problem rather than by method, and the partition is not exhaustive: it reflects how the National Library of Medicine indexes the literature, not a claim that all of psychophysics reduces to these two headings. Vision, touch, taste, and olfaction each host their own psychophysical traditions that the MeSH tree distributes elsewhere. The subtypes are also orthogonal to the discipline's methods, since signal detection procedures and adaptive staircases apply across every sensory domain alike.

Both subtypes remain plain text above because neither has its own pillar page yet. The site nonetheless treats signal detection as a first-class topic on its signal detection theory page, and its psychoacoustic cousins appear under pitch perception, auditory perception, and sound localization.

Weber's Law and Fechner's Law

Ernst Heinrich Weber, working on the sense of touch and the discrimination of lifted weights, observed that the just-noticeable difference between two stimuli is not a fixed physical amount but a constant proportion of the standard (Ross & Murray, 1996). Doubling the standard doubles the increment needed to detect a change. Formally, the ratio of the difference threshold to the baseline intensity is constant, and that constant is the Weber fraction. A Weber fraction of 0.02 for lifted weight means a 2 percent change is the smallest reliably detectable one at any baseline, so 2 grams against 100 grams and 10 grams against 500 grams are equally discriminable.

Fechner took Weber's empirical regularity and integrated it into a scale of sensation. If every just-noticeable difference is treated as a subjectively equal step, and each such step corresponds to a constant fractional increase in intensity, then sensation grows as the logarithm of intensity (Fechner, 1860/1966). Fechner's logarithmic law predicts diminishing returns: as intensity climbs, ever larger physical increases are needed to produce the same felt increase, which is why a candle brightens a dark room dramatically but adds little to a sunlit one. The law holds well across the middle of most intensity ranges and breaks down near absolute threshold and at very high intensities.

Stevens' Power Law and the Scales of Measurement

S. S. Stevens challenged Fechner's logarithm on both empirical and logical grounds. Using magnitude estimation, in which observers assign numbers directly proportional to the perceived intensity of a stimulus, Stevens found that sensation grows not as the logarithm of intensity but as a power function of it (Stevens, 1957). Perceived magnitude equals intensity raised to an exponent that is characteristic of the continuum: roughly 0.33 for brightness, near 1.0 for apparent line length, about 1.45 for lifted weight, and as high as 3.5 for electric shock. An exponent below one describes a compressive sense that grows more slowly than its stimulus; an exponent above one describes an expansive sense that grows faster. Stevens argued that Fechner had mistaken the discriminability of a continuum for its felt magnitude, and he pressed the case in a pointed reappraisal of the founder's program (Stevens, 1961).

Underlying the dispute was a deeper contribution: Stevens' taxonomy of measurement scales. He classified any measurement as nominal, ordinal, interval, or ratio according to the mathematical transformations that leave its meaning intact, and argued that magnitude estimation yields a ratio scale of sensation whereas the counting of just-noticeable differences yields only an interval scale (Stevens, 1946). The framework reshaped how psychology reasoned about permissible statistics. A related scaling tradition descends from L. L. Thurstone, whose law of comparative judgment modeled each stimulus as a distribution of momentary sensory values and derived an interval scale from the proportion of times one stimulus is judged greater than another (Thurstone, 1927). Thurstonian scaling anticipated signal detection theory by locating discrimination in overlapping internal distributions rather than a fixed threshold.

The Classical Psychophysical Methods

Fechner bequeathed three procedures for estimating thresholds, and they remain the reference points against which modern methods are judged (Gescheider, 1997). In the method of limits, the experimenter raises or lowers intensity in small steps until the observer's report flips, and the threshold is the average crossing point. In the method of constant stimuli, a fixed set of intensities is presented many times in random order, and the threshold is read off the resulting psychometric function. In the method of adjustment, the observer controls the stimulus directly and sets it to the point of subjective equality with a standard. Each trades efficiency against bias in a different way: limits is fast but prone to anticipation and habituation errors, constant stimuli is unbiased but wasteful, and adjustment is quick but confounds perception with motor control. A separate refinement concerns the form of the response rather than the placement of trials: in a forced-choice design the observer must say which of two or more intervals held the signal rather than simply reporting yes or no, which largely removes the criterion contamination that afflicts yes/no reports and yields a threshold less entangled with the observer's caution (Kingdom & Prins, 2016).

Because fixed grids of intensity spend most trials far from the threshold, where they are uninformative, adaptive methods place trials where the data are most useful. The staircase procedure moves intensity up after an incorrect response and down after correct ones, converging on the intensity that yields a target performance level (Cornsweet, 1962). Bayesian adaptive methods go further, maintaining a running probability distribution over the threshold and choosing each stimulus to sharpen that estimate fastest, the strategy embodied in the QUEST algorithm (Watson & Pelli, 1983). Whatever method delivers the data, the threshold itself is extracted by fitting a psychometric function, and the statistics of that fit have received their own rigorous treatment (Wichmann & Hill, 2001).

Signal Detection Theory

The classical threshold assumes a sharp boundary between the sensed and the unsensed, but detection near threshold is noisy, and an observer's willingness to say yes depends on more than the stimulus. Signal detection theory, imported into psychophysics from radar engineering, replaces the threshold with two independent quantities (Tanner & Swets, 1954). Sensitivity, indexed by d-prime, measures how far the internal response to a signal separates from the response to noise alone, in units of their shared standard deviation. The response criterion measures where the observer places the yes/no boundary, a decisional variable governed by expectations and payoffs rather than by the senses. A cautious observer and a liberal one can share identical sensitivity yet produce very different hit and false-alarm rates.

Plotting the hit rate against the false-alarm rate as the criterion varies traces the receiver operating characteristic, whose shape reveals sensitivity independent of bias (Green & Swets, 1966). The framework transformed perception research and spread far beyond it, into recognition memory, diagnostic radiology, and any decision making task in which a faint signal must be judged against uncertainty. Its intellectual history proved longer and more tangled than the standard account allowed, with roots that a recent reappraisal traces back well before its 1950s psychophysical debut (Wixted, 2020).

Worked Example

Consider a weight-discrimination experiment. An observer lifts a standard of 100 grams and can reliably tell it from a 102-gram comparison but not from a 101-gram one, so the just-noticeable difference is about 2 grams. The Weber fraction is therefore the difference threshold divided by the standard: 2 divided by 100, or 0.02. Weber's law then predicts the difference threshold at any other baseline by multiplication. At a 500-gram standard the predicted just-noticeable difference is 0.02 times 500, which equals 10 grams, five times the increment needed at 100 grams even though the fraction is unchanged.

The two scaling laws diverge sharply on the same continuum. Doubling the stimulus from 100 to 200 grams adds a constant increment under Fechner's logarithm, because the logarithm of 200 over 100 is the logarithm of 2 regardless of where on the scale the doubling occurs. Under Stevens' power law with the lifted-weight exponent of 1.45, the same doubling multiplies perceived heaviness by 2 raised to the 1.45 power, which is about 2.73, so the sensation nearly triples. The disagreement is empirical and testable, which is precisely why magnitude estimation could adjudicate it.

Now score the same detection data with signal detection theory. Suppose that on signal-present trials the observer says yes 93.3 percent of the time (the hit rate) and on signal-absent trials says yes 6.7 percent of the time (the false-alarm rate). Converting each proportion to a z-score gives +1.5 for the hits and -1.5 for the false alarms. Sensitivity is the difference of these z-scores, d-prime equals 1.5 minus negative 1.5, which is 3.0. The criterion is the negative average of the two z-scores, which here is zero, marking an unbiased observer who favors neither response. A shift in payoffs that made misses costly would move the criterion without touching d-prime, exactly the dissociation the theory was built to expose.

Discussion

Psychophysics matters because it is the metrological foundation of the mind sciences. Every claim that an observer perceives a difference, adapts to a background, or improves with practice ultimately rests on a threshold or a scaling estimate, and the credibility of that claim is only as good as the psychophysical method behind it. This is why the field's apparent disputes are so consequential. Whether sensation grows as a logarithm or a power function is not a quibble over curve fitting; it determines how perceptual data may legitimately be summed, averaged, and compared. Whether detection reflects a threshold or a criterion decides whether a training gain reflects sharper senses or merely a bolder decision rule.

The signal detection revolution did not so much refute the classical laws as reframe them, showing that what looked like a sensory limit often hid a decisional one. Modern practice keeps both traditions: Weber fractions and scaling exponents still summarize the operating characteristics of a sense, while d-prime and the receiver operating characteristic protect those summaries from contamination by bias. The enduring lesson is that the observer is not a passive meter but an active decider, and good psychophysics is the discipline of measuring the sense without being fooled by the decision.

Current Directions

Three currents define contemporary psychophysics. The first is statistical: estimating a psychometric function well from limited, noisy, sometimes overdispersed data is harder than the classical fitting recipes assumed, and Bayesian methods now deliver threshold and slope estimates with honest uncertainty and built-in robustness to lapses (Schütt et al., 2016). The second is neural: the century-old Weber's law has been given a mechanistic foundation in the dynamics of cortical circuits, showing how a constant fractional threshold can emerge from the way populations of neurons accumulate sensory evidence over time (Pardo-Vázquez et al., 2019). The third is historiographic and conceptual: a careful reexamination of signal detection theory has recovered a forgotten lineage and clarified what the framework does and does not assume, with direct consequences for how sensitivity is measured in memory and perception alike (Wixted, 2020). Comprehensive modern treatments now integrate these strands into a single practical toolkit for the working experimenter (Kingdom & Prins, 2016).

Common Misconceptions

Weber's law says the just-noticeable difference is a fixed physical amount.

It says the opposite. The just-noticeable difference grows in proportion to the baseline intensity; what stays constant is the ratio, the Weber fraction, not the absolute increment (Ross & Murray, 1996).

A threshold is a fixed intensity below which nothing is ever perceived.

Detection probability rises gradually, so a threshold is a statistical construct: the intensity yielding a chosen criterion level of performance, read from a psychometric function rather than a hard cutoff (Gescheider, 1997).

A high false-alarm rate simply means poor perception.

Signal detection theory shows that hit and false-alarm rates confound sensitivity with the response criterion. A liberal observer can post many false alarms while sensing perfectly well; only d-prime isolates the perceptual component (Green & Swets, 1966).

Key Researchers

Gustav Fechner (1801-1887). Physicist and philosopher at Leipzig who founded psychophysics with the Elemente der Psychophysik (1860), deriving the logarithmic law and formalizing the classical methods. Wikipedia

Frederick Kingdom. Vision scientist at McGill University whose widely used practical textbook codifies modern psychophysical method; an ORCID was individually searched and none was found on his faculty or laboratory pages, a genuine absence rather than an omission. Faculty · Google Scholar

Denis Pelli. Perceptual scientist at New York University and co-author of the QUEST adaptive Bayesian method, with continuing work on the limits of visual detection and reading. ORCID 0000-0001-7100-5659 · Faculty

Stanley Smith Stevens (1906-1973). Harvard psychophysicist who introduced magnitude estimation, the power law of sensation, and the nominal-ordinal-interval-ratio taxonomy of measurement scales. Wikipedia

Louis Leon Thurstone (1887-1955). Pioneer of psychological scaling whose law of comparative judgment modeled discrimination through overlapping internal distributions, anticipating signal detection theory. Wikipedia

Ernst Heinrich Weber (1795-1878). Leipzig physiologist whose studies of touch and lifted weights established that the difference threshold is a constant fraction of intensity, the empirical law Fechner later integrated. Wikipedia

Felix Wichmann. Computational perception scientist at the University of Tübingen whose work set the standard for rigorous psychometric-function fitting and Bayesian estimation. ORCID 0000-0002-2592-634X · Google Scholar

Glossary

Absolute Threshold.

The smallest stimulus intensity an observer can reliably detect against a null background.

Adaptive Method.

A procedure that chooses each trial's intensity from prior responses so as to concentrate testing near the threshold.

Difference Threshold.

The smallest change in intensity that can be told apart from a standard; also called the just-noticeable difference.

Forced-Choice.

A response format in which the observer selects which of two or more intervals contained the signal, reducing the influence of response bias on the threshold estimate.

Just-Noticeable Difference.

The increment in stimulus intensity that an observer detects on a criterion proportion of trials, Fechner's unit of sensation.

Method of Adjustment.

A classical procedure in which the observer directly controls the stimulus and sets it to a target, such as the point of subjective equality.

Method of Constant Stimuli.

A classical procedure presenting a fixed set of intensities in random order to trace the full psychometric function.

Method of Limits.

A classical procedure that raises or lowers intensity in steps until the observer's report changes, averaging the crossing points.

Point of Subjective Equality.

The comparison intensity that an observer judges equal to a standard, marking the center of a discrimination function.

Power Law.

Stevens' account in which perceived magnitude equals stimulus intensity raised to a continuum-specific exponent.

Psychometric Function.

The curve relating the probability of a given response to stimulus intensity, from which threshold and slope are estimated.

Receiver Operating Characteristic.

A plot of hit rate against false-alarm rate across criteria that reveals sensitivity independent of response bias.

Response Criterion.

The internal cutoff at which an observer decides to respond yes, a decisional variable set by expectations and payoffs.

Sensitivity.

The separation between the internal responses to signal and to noise, indexed by d-prime, independent of the criterion.

Signal Detection Theory.

The framework that models detection as a decision under uncertainty, dissociating perceptual sensitivity from response criterion.

Staircase Procedure.

An adaptive method that lowers intensity after correct responses and raises it after errors to converge on a target performance level.

Weber Fraction.

The constant ratio of the difference threshold to the baseline intensity that characterizes a sensory continuum.

Weber's Law.

The principle that the just-noticeable difference is a constant proportion of the standard intensity.

Frequently Asked Questions

What is psychophysics in simple terms?
Psychophysics is the science that measures how physical stimuli map onto perceived sensations, using thresholds and scaling functions estimated from an observer's responses (Fechner, 1860/1966).

What is the difference between the absolute and difference thresholds?
The absolute threshold is the faintest intensity detectable against nothing, whereas the difference threshold is the smallest detectable change from an existing standard (Gescheider, 1997).

What does Weber's law state?
It states that the just-noticeable difference is a constant fraction of the baseline intensity, so larger standards require proportionally larger increments to be told apart (Ross & Murray, 1996).

How does Fechner's law differ from Stevens' power law?
Fechner held that sensation grows as the logarithm of intensity, while Stevens found from magnitude estimation that it grows as a power function whose exponent varies by sense (Stevens, 1957).

Why did Stevens want to repeal Fechner's law?
Stevens argued that counting just-noticeable differences yields only an interval scale of discriminability, not the ratio scale of felt magnitude that direct estimation reveals (Stevens, 1961).

What are the classical psychophysical methods?
They are the method of limits, the method of constant stimuli, and the method of adjustment, three procedures Fechner devised to estimate thresholds from behavior (Gescheider, 1997).

What is an adaptive staircase?
A staircase raises intensity after errors and lowers it after correct responses, converging efficiently on the intensity that yields a target level of performance (Cornsweet, 1962).

How does signal detection theory improve on the threshold concept?
It replaces a single threshold with two independent measures, sensitivity and criterion, so that perceptual acuity can be separated from an observer's willingness to respond (Green & Swets, 1966).

References