eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

The Gold Standard of Psychophysics

—

by

eduKate Asia
Three learners review open books together at a classroom table, with stacks of textbooks, stationery and a whiteboard in the bright room.

Two students listen to a faint tone. One says ‘yes’ almost every time, while the other reports hearing it only when confident. Which student has better hearing? More yes answers do not establish higher sensitivity: response bias, false alarms and uncertainty matter. The gold standard of psychophysics separates the physical stimulus, perceptual performance and the observer’s decision strategy.

Psychophysics studies how measurable properties of light, sound, touch and other stimuli relate to perception. It uses carefully designed discrimination, detection and comparison experiments and models the probability of a response. A psychophysical threshold is not an absolute property extracted from one sensation. It is a defined criterion under specific equipment, task and statistical assumptions.

A 2026 review of psychophysics and sensation measurement revisits Weber, Fechner and modern stochastic thresholds. Research on psychophysical experiment design and adaptive procedures and lapses explains how trial choices shape sensitivity estimates. This O–P root complements eduKate’s Psychology, Acoustics and Optics articles.

Quick information: a reported threshold has a method and uncertainty

A credible psychophysical experiment specifies a calibrated stimulus, trial sequence, task, response choices, chance level, threshold criterion, participant population, number of trials, omissions and model. Different detection and discrimination tasks can produce different thresholds without one necessarily being defective. A fitted curve needs raw response counts and uncertainty. A research measurement is not automatically a medical diagnosis or a judgment about an individual’s ability.

Chapter index

Worked case one: a 75 percent criterion

In a fictional two-alternative visual task, a participant is correct on 12 of 20 trials (60%) when contrast is 0.10, and 16 of 20 (80%) when contrast is 0.20. If the investigator uses 75% correct as the predefined criterion and makes a simple linear interpolation between these two observed proportions, the approximate crossing occurs at 0.175 contrast units. It is only an illustrative interpolation, not a fitted psychometric threshold.

With only twenty trials per contrast, the true response probabilities are uncertain, and the psychometric function may not be linear. An actual analysis would use a defensible psychometric model, suitable stimulus levels, lapse considerations and confidence or credible intervals. The example demonstrates why a threshold is linked to both measurement and a declared criterion.

Twenty controls for reliable psychophysics

1. Define the physical stimulus and the response task

Psychophysics investigates relations between measurable stimuli and perceptual judgements. A person may detect a faint tone, distinguish two textures, report brightness or decide whether a moving pattern goes left or right. These tasks measure different aspects of processing. Define the stimulus property, equipment, presentation, alternatives, correct response and target variable before interpreting a threshold. A laboratory device can produce precise physical intensity while the person’s judgement still varies. The goal is an interpretable relation between stimulus and behaviour, not a claim that perception is perfectly objective.

2. Distinguish detection from discrimination

Detection asks whether a signal is present against a background; discrimination asks whether two presented stimuli differ. A participant can identify the stronger of two tones even when neither would be described as readily heard alone in another task. Thresholds depend on the procedure and its chance response baseline. Do not compare a yes/no detection score directly with a two-alternative forced-choice score without accounting for that distinction. Describe the number of possible answers, feedback, and which quantity is used as the threshold criterion.

3. Use calibrated stimulus intensity and time

Contrast, sound pressure, vibration strength or luminance requires a physical definition and appropriate instrument calibration. A monitor’s brightness setting is not automatically a calibrated luminance, and a computer’s volume slider does not directly describe the sound at the listener’s ear. Record the stimulus geometry, duration, adaptation conditions, viewing or listening distance and ambient environment as needed. Instrument drift can create apparent perceptual changes between sessions. A physical input must be reliable before a sensory threshold can be interpreted as human variability.

4. Recognise that sensory judgements are probabilistic

An observer may notice the same weak stimulus on some trials and miss it on others because of internal variation, attention, noise and decision strategies. A psychometric function models the probability of a defined response across stimulus values rather than an exact binary cut-off in every trial. The 2026 measurement review traces how the concept of psychophysical threshold developed from classical step-like ideas to stochastic models. A single trial does not uniquely identify sensitivity. Repeat observations are needed, with their uncertainty and task conditions visible.

5. Define the threshold criterion before testing

In a two-alternative forced-choice discrimination task, 75% correct is a common threshold criterion between 50% chance and near-perfect performance, but other tasks or purposes can use different conventions. The review of children’s psychophysical measurement explains why task and criterion must be specified. Do not simply call the weakest presented stimulus ‘the threshold’ after seeing data. State whether the threshold is an absolute, difference or model-specific criterion and how it was estimated.

6. Use a just-noticeable difference cautiously

A just-noticeable difference describes a small stimulus change detectable under a stated probability and task. It depends on sensory modality, baseline intensity, adaptation and methodological choices. The historical Weber relation connects some difference thresholds to reference intensity over limited ranges, not every stimulus and perception. A threshold at one sound level cannot be transferred automatically to another or another person. Show the original reference stimulus, comparison values and criterion when presenting a JND.

7. Treat Weber–Fechner relations as models, not universal laws

Weber’s law approximately describes proportional difference thresholds in suitable conditions; Fechner’s historical work linked such differences to models of sensation magnitude. The 2026 review of psychophysical measurement explains the theoretical assumptions behind that development. Real sensory systems have limits, adaptation and nonlinearities that challenge simple generalisations. The ratio ΔI/I should be reported as an empirical or stipulated approximation for the particular modality and range, not a universal rule defining human experience.

8. Distinguish sensitivity from response bias

A person may have the ability to detect a weak signal but adopt a conservative criterion for saying ‘present’. Another may respond yes frequently and produce more hits and false alarms. Signal detection theory helps separate performance sensitivity from response policy under specified distribution assumptions. The psychometric decision-making review describes how choice bias affects measured functions. A high percentage of yes responses does not demonstrate superior hearing or vision. Record both signal-present and signal-absent trials when the task requires them.

9. Calculate hit and false-alarm rates correctly

A hit is a signal-present trial answered as present; a false alarm is a signal-absent trial answered as present. Divide each by its own eligible trial count. If 80 of 100 signal trials are detected, hit rate is 80%; if 20 of 100 noise-only trials are answered yes, false-alarm rate is 20%. Comparing these rates helps distinguish genuine sensitivity from willingness to say yes. Handle extreme values and missed or invalid trials using a documented method rather than silently changing denominators.

10. Model lapses and omitted responses

Even a clearly detectable stimulus may be answered incorrectly because a participant blinks, becomes distracted or presses the wrong key. These lapses can distort estimates of psychometric thresholds and slope. The 2018 adaptive psychometric-function study explicitly models lapses and omitted responses. A fitted curve that assumes perfect behaviour can misinterpret occasional mistakes as weak sensory sensitivity. Do not discard all difficult trials or only the inconvenient outliers. Define how omissions and low-engagement responses enter analysis.

11. Control trial order, anticipation and learning

Performance can change because stimuli become predictable, the participant adapts, or practice improves task familiarity. Randomisation and counterbalancing can reduce some bias, but time effects still need checking. A 2020 two-alternative forced-choice study found that presentation order can distort adaptive threshold estimation. A threshold collected when the reference always appears first is not necessarily equivalent to one from a counterbalanced procedure. Record sequence, rest breaks, feedback and any task training before claiming a stable individual property.

12. Choose adaptive methods with explicit assumptions

Staircase and Bayesian adaptive procedures concentrate measurements near informative stimulus levels and may estimate a threshold with fewer trials. However, a method optimised for one smooth monotonic psychometric function can fail when observers lapse, have non-monotonic performance or exhibit order bias. The 2026 comparison of threshold-estimation methods demonstrates advantages and limitations under different simulated conditions. A shorter procedure is not automatically a more accurate one. State selection rule, stopping criterion, model and validation.

13. Use sufficient trials and report uncertainty

Twenty trials at one intensity give a noisier estimate of true response probability than an adequately larger independent set, all else equal. Confidence intervals or credible intervals should reflect trial count, dependence and model assumptions. An exact threshold to several decimal places may be unjustified when the data contain a handful of near-threshold judgments. Report number of valid trials by stimulus and participant, excluded observations and the scope of inference. A smooth fitted curve does not create observations that were never collected.

14. Distinguish threshold, slope and upper asymptote

A psychometric function can shift horizontally, change steepness or approach an upper performance limit below 100%. Those changes can reflect different underlying processes. A threshold change is not necessarily the same as an improvement in discrimination precision, and a lower asymptote can reflect lapses. The review of psychometric-function analysis describes how different experimental designs help distinguish these characteristics. Present model parameters with units, uncertainties and supporting data rather than treating one summary number as the entire sensory response.

15. Measure participants rather than multiplying trials into people

A thousand judgements from one participant do not create a thousand independent participants. Within-observer variation and between-observer variation answer different questions. If conclusions concern a population, sample suitable individuals and use analysis respecting repeated trials nested within people. A psychophysical curve can characterise one well-trained observer without establishing how children, older adults or other communities respond. State the study population and avoid extending individual findings to everyone.

16. Adapt tasks to children and accessibility needs

Instruction comprehension, language, motor responses, attention, fatigue and sensory or cognitive differences can influence performance. The psychophysical measurement review for children details how practice, feedback and lapses complicate interpretation. A task that a child does not understand cannot provide a reliable sensory threshold merely because the computer calculated one. Use appropriate accessible response formats, monitor engagement and avoid treating a limitation of the test as a deficiency of the person.

17. Separate laboratory perceptual performance from clinical diagnosis

Psychophysical tools contribute to vision, hearing and sensory research, and some clinical tests use similar principles. But a research contrast threshold or tone-discrimination score is not independently a diagnosis of eye, ear or neurological disease. Clinical interpretation requires qualified practitioners, standardised instruments, symptoms and relevant medical evidence. Do not use an experimental result to self-diagnose or decide treatment. Describe the purpose and validation of the specific test before making medical or educational recommendations.

18. Evaluate context, adaptation and cross-modal effects

Sensory judgments can depend on surrounding light or sound, recent exposure, attention and expectations. A contrast visible on a neutral background may be harder to detect when the background changes. Adaptive effects and multisensory cues can alter performance. Report relevant context and control extraneous cues where the purpose is to isolate one property. If the study involves complex scenes, a simple intensity–response function may not capture every mechanism. Explain what was controlled and what remains part of naturalistic perception.

19. Protect privacy and fair use of perceptual measurements

Sensory performance can reveal information about health or disability and should be collected and stored responsibly. Obtain appropriate consent, protect participant identification and explain how data may be reused. Do not turn perceptual differences into judgments of intelligence, honesty, employability or human worth. A sensory threshold is a performance measure under test conditions, not a global character assessment. When adapting tests to different populations, assess validity and access rather than assuming the same interface suits everyone.

20. Make the full experiment reconstructable

A dependable report includes physical stimulus calibration, timing, task instructions, response alternatives, randomisation, participants, trial counts, exclusion rules, threshold criterion, model and uncertainty. Describe both successful fits and non-ideal behaviour that challenged the assumptions. If a new method changes the threshold estimate, investigate why instead of selecting whichever result looks best. The gold standard of psychophysics is a transparent quantitative account of perception that another researcher can reproduce and critically evaluate.

Worked case two: hit rate and false alarms

In a fictional yes/no tone-detection experiment, 80 of 100 signal-present trials receive a yes response, while 20 of 100 noise-only trials also receive yes. The hit rate is 80% and the false-alarm rate is 20%. These rates give a more informative description than saying the participant responded yes on 100 of 200 trials. A signal-detection analysis could use both rates to model sensitivity and criterion under appropriate assumptions.

If the equal-variance Gaussian signal detection model applies, the illustrative d-prime is z(0.80) − z(0.20) ≈ 1.68. The model assumption and sampling error matter, and this is not a clinical hearing result. The numbers are invented. A second person with the same hit rate but higher false-alarm rate would not necessarily have the same inferred sensitivity.

Third worked case: a Weber fraction is not a universal human constant

Suppose a fictional difference-detection task follows an approximate Weber fraction ΔI/I = 0.10 within the tested range. Against a reference intensity of 100 units, an increment of 10 units meets the stipulated criterion. Against 200 units, the equivalent ratio requires an increment of 20 units. The same absolute increase does not have the same relative size.

This arithmetic is a conditional example, not a claim that Weber’s law governs all hearing or vision tasks. Near extremes or under adaptation, the relationship can differ. The experiment must specify the stimulus, baseline and measurement criterion before interpreting a just-noticeable difference.

Psychophysics quality checklist

  • What physical stimulus and sensory task were actually measured?
  • Were equipment intensity and presentation timing calibrated?
  • Is the task detection, discrimination or magnitude estimation?
  • What is the chance response baseline and threshold criterion?
  • How many valid trials were collected at each stimulus level?
  • Were lapses, omissions and response bias considered?
  • Could trial order, learning or adaptation distort performance?
  • Does the statistical model match the observed behaviour?
  • Are sensitivity, slope and upper performance limit distinguished?
  • Are participant-level and trial-level observations analysed appropriately?
  • Does the interpretation avoid unwarranted clinical or social judgments?

Frequently asked questions

What is psychophysics?

The scientific study of how physical stimulus properties relate to reported or measured perceptual responses.

What is a psychometric function?

A relationship between stimulus characteristics and the probability of a defined behavioural response.

What is a sensory threshold?

A criterion-based level of stimulus at which the observer reaches specified performance under a particular task.

What is a just-noticeable difference?

A small change in stimulus that can be detected or discriminated under a declared criterion and context.

Is Weber’s law true for all stimuli?

No. It is an approximate relationship under some ranges and modalities.

What is a two-alternative forced-choice task?

A task in which a participant selects one of two alternatives, with a defined chance level often of 50%.

Why is 75% correct sometimes a threshold?

It is a conventional performance criterion in some two-alternative tasks, not a universal definition.

What is a lapse rate?

An estimate of occasional responses unrelated to the stimulus, which may reflect attention or response errors.

What is the difference between detection and response bias?

Detection concerns distinguishing signal from noise; response bias reflects how willing the participant is to select one answer.

Are adaptive staircases always more accurate?

No. Efficiency depends on the model, actual participant behaviour and absence of unaccounted order or lapse effects.

Does a research sensory threshold diagnose disease?

Not automatically. Clinical diagnosis requires standardised testing and qualified professional interpretation.

Can different participants have different psychometric functions?

Yes. Performance and task understanding can vary within and between individuals and conditions.

Research references and educational limitations

All participants, response numbers and intensity values in the examples are fictional. This article explains scientific measurement and does not offer a clinical diagnosis, official sensory screening result or proof of innate differences among people. The Gold Standard is eduKate’s editorial standard for reproducible and responsible research.