Contrast Sensitivity

Abstract

Contrast sensitivity, which the Medical Subject Headings vocabulary classifies under visual acuity, is the visual system's ability to detect differences in luminance between a feature and its background. It is measured not at a single size but across a range of spatial frequencies, yielding the contrast sensitivity function (CSF): an inverted-U curve that peaks for medium patterns and falls off for both coarse and fine ones. This article treats contrast sensitivity as the fundamental currency of spatial vision rather than a clinical afterthought: how contrast and the CSF are defined and measured, why the evidence points to a bank of spatial-frequency-tuned channels rather than one detector, how the retina and cortex encode contrast, and how modern adaptive methods estimate the whole function in minutes. It covers gratings, Fourier analysis, selective adaptation, and the neural substrate.

Keywords: contrast sensitivity function, spatial frequency, spatial vision

What Contrast Sensitivity Is

Contrast sensitivity is the ability to see a faint difference in luminance. Visual acuity, the familiar quantity measured by a letter chart, asks how small a high-contrast target can be and still be resolved; contrast sensitivity asks a more general question: how little contrast a pattern of a given size needs before it becomes visible at all. Acuity is one point on the contrast sensitivity function, the single spatial frequency at which a very high-contrast target drops below threshold. The full function describes vision across every scale, which is why it is the more complete description of spatial vision and why the Medical Subject Headings vocabulary files contrast sensitivity beneath visual acuity rather than treating the two as the same thing.

The stimulus that made contrast sensitivity measurable is the sinusoidal grating: a pattern whose luminance varies as a sine wave across space, defined by just two numbers. Its spatial frequency, given in cycles per degree of visual angle, fixes how coarse or fine the pattern is, and its contrast fixes how deep the light-to-dark modulation is. Contrast is quantified by the Michelson formula, the difference between the maximum and minimum luminance divided by their sum, so that a faint grating has a contrast near zero and a crisp black-and-white one a contrast near one. The great advantage of the grating is that any image can be analysed as a sum of such gratings, so the visibility of gratings characterises the visibility of everything.

Keeping the quantities of contrast vision apart matters, because they are easily conflated and each answers a different question. Table 1 lays out the principal terms: the physical measure of stimulus contrast, the scale parameter of the pattern, the behavioural threshold, and the sensitivity that is its reciprocal.

Measuring Contrast Sensitivity

To measure the function, an observer is shown gratings at a range of spatial frequencies, and at each frequency the contrast is varied until it reaches the threshold where the grating is just detectable. The reciprocal of that threshold is the sensitivity at that frequency, and the set of sensitivities across frequencies is the CSF. John Robson first charted the function this way, measuring sensitivity jointly across spatial and temporal frequency and showing that the spatial curve is not monotonic but band-pass, peaking at intermediate frequencies (Robson, 1966). The GratingContrastDemo renders a single grating whose spatial frequency and Michelson contrast the reader can set, with a model threshold marking where it should disappear.

The shape of the human CSF is robust. Sensitivity is highest for medium patterns, around two to five cycles per degree, and falls away in both directions: toward low frequencies because of lateral inhibition in the retina, which suppresses responses to broad, uniform regions, and toward high frequencies because of the optical blur of the eye and the finite spacing of the photoreceptors. The ContrastSensitivityFunctionDemo plots this inverted-U on logarithmic axes and lets the reader move the peak sensitivity, the peak frequency, and the bandwidth, reading off the high-frequency cutoff that corresponds to the acuity limit.

The function is also a sensitive index of visual health, which has driven a long effort to measure it efficiently. Classical methods estimated one threshold at a time with an adaptive staircase; the QUEST procedure made each trial count by maintaining a running Bayesian estimate of the threshold and placing the next trial where it is most informative (Watson & Pelli, 1983). The decisive step for clinical use was to estimate the whole function rather than isolated points: the quick CSF method treats the four parameters of the CSF as the thing to be estimated and chooses each grating to maximise the expected information about all of them at once, recovering the entire curve in a few minutes (Lesmes, Lu, Baek, & Albright, 2010). The same Bayesian logic generalises to any psychometric experiment in the QUEST+ framework (Watson, 2017), and the modern rationale for measuring contrast sensitivity, including why the function rather than acuity alone should be the clinical target, has been set out directly (Pelli & Bex, 2013). These adaptive methods are now used to track contrast sensitivity in low-vision populations, where the quick CSF relates closely to other measures of functional vision (Stalin & Dalton, 2020).

Spatial-Frequency Channels

The single most consequential discovery about contrast sensitivity is that the function is not the signature of one detector but the envelope of many. Fergus Campbell and John Robson proposed that the visual system analyses an image not as a whole but through a bank of channels, each tuned to a narrow band of spatial frequency, by applying Fourier analysis to the visibility of gratings: a complex grating made of several frequencies became visible exactly when its most sensitive component reached that component's own threshold, as if the frequencies were detected independently (Campbell & Robson, 1968). A single broadband detector could not behave this way.

The channels were confirmed by selective adaptation. Colin Blakemore and Fergus Campbell showed that staring at a high-contrast grating of one spatial frequency temporarily raises the threshold for gratings near that frequency, while leaving sensitivity to very different frequencies untouched, carving a localised notch out of the CSF and revealing the bandwidth of the adapted channel (Blakemore & Campbell, 1969). Converging evidence came from summation at threshold: Norma Graham and Jacob Nachmias reasoned that if two frequencies feed one detector, combining them should lower the threshold more than if each has its own channel, and the data matched the multiple-channels prediction, not the single-channel one (Graham & Nachmias, 1971). The ChannelBankDemo draws a set of overlapping channel sensitivity curves whose upper envelope is the CSF, and lets the reader adapt a channel to see the notch appear.

Two refinements complete the picture. Michael Georgeson and George Sullivan found that although threshold contrast varies strongly with spatial frequency, the perceived contrast of a supra-threshold grating is nearly constant across frequencies, a phenomenon they called contrast constancy and attributed to the channels deblurring the retinal image by compensating for their own differing sensitivities (Georgeson & Sullivan, 1975). And the channels interact: a masking grating raises the threshold for a test grating in a way that depends on their relative frequency and contrast, a dependence Gordon Legge and John Foley captured with a model of contrast masking that remains a template for how channel responses are combined and normalised (Legge & Foley, 1980). Sensitivity also varies across the visual field, falling with eccentricity in a lawful way that probability summation over independent regions helps describe (Robson & Graham, 1981).

Neural Encoding of Contrast

The channels are not a mathematical convenience; they have a physiological substrate that begins in the retina. Christina Enroth-Cugell and John Robson measured the contrast sensitivity of individual retinal ganglion cells in the cat and found that each responds to contrast within a limited band of spatial frequencies set by the antagonistic centre-surround structure of its receptive field; they also drew the lasting distinction between linear (X) and nonlinear (Y) cells, establishing that contrast and spatial-frequency tuning are built into the very first stage of the visual pathway (Enroth-Cugell & Robson, 1966). The low-frequency falloff of the CSF is already present here, in the surround's suppression of responses to uniform fields.

In the cortex the tuning sharpens. Russell De Valois, Duane Albrecht, and Lisa Thorell recorded cells in macaque visual cortex and found them selective for spatial frequency, each responding best to a narrow band and collectively tiling the range of frequencies the animal can see, the single-neuron counterpart of the psychophysical channels (De Valois, Albrecht, & Thorell, 1982). Blakemore and Campbell had already predicted such neurons from the adaptation data, inferring the existence of cortical units selectively sensitive to the size and orientation of retinal images before they were directly recorded (Blakemore & Campbell, 1969). To connect the physiology back to behaviour, Andrew Watson and Albert Ahumada built a standard model observer that predicts human foveal contrast detection from a bank of frequency- and orientation-tuned filters followed by a decision stage, formalising how the neural channels produce the measured CSF (Watson & Ahumada, 2005). Figure 1 traces the path from the grating to the perceptual threshold.

Figure 1

Worked Example: Reading Thresholds Off the CSF

The relationship between sensitivity and threshold is a reciprocal, so a CSF can be read directly as a map of how much contrast each pattern needs. Model the function as a log-Gaussian, a convenient and widely used form, with sensitivity

S(f) = S_max · exp[ −(ln f − ln f_peak)² / (2σ²) ]

where f is spatial frequency in cycles per degree. Take a representative healthy observer with peak sensitivity S_max = 300 at a peak frequency f_peak = 4 cycles per degree, and a bandwidth σ = 0.75 in natural-log units (a full width at half maximum of about 2.5 octaves, the typical human value).

At the peak, f = 4, the exponent is zero, so S = 300 and the contrast threshold is 1 / 300 = 0.0033, that is 0.33 percent contrast — the faintest 4-cycle grating this observer can see. Move two octaves away in either direction, to 1 or 16 cycles per degree. The log-frequency distance is ln(16/4) = ln 4 = 1.386, so the exponent is −(1.386)² / (2 × 0.75²) = −1.922 / 1.125 = −1.708, giving S = 300 × e^−1.708 = 300 × 0.181 = 54.4 and a threshold of 1 / 54.4 = 1.84 percent contrast. Sensitivity has fallen more than fivefold, so a grating five times fainter is needed at the peak than two octaves out.

The high-frequency cutoff — the acuity limit — is where sensitivity drops to 1, the point at which even full (100 percent) contrast is just barely detectable. Setting S = 1 gives (ln f − ln 4)² = 2σ² ln S_max = 2 × 0.5625 × ln 300 = 1.125 × 5.704 = 6.42, so ln f − ln 4 = 2.53 and f = 4 × e^2.53 = 50 cycles per degree. That cutoff is the quantity a letter chart reports as acuity: roughly 50 cycles per degree corresponds to the fine end of normal vision. The GratingContrastDemo and ContrastSensitivityFunctionDemo use exactly this model, so these numbers can be read off the curve directly. The worked example makes the article's central claim concrete: acuity is a single point — the high-frequency intercept — on a function that describes vision at every scale.

Discussion

Contrast sensitivity reframed spatial vision. Before the CSF, visual resolution was summarised by a single acuity number; after it, vision was understood as a profile across spatial scales, and the Fourier grating became the probe that made that profile measurable. The deepest idea to emerge was the multiple-channels hypothesis: the finding that the smooth CSF is the envelope of many narrowly tuned detectors, established by adaptation and summation and then confirmed in the tuning of single cortical neurons. That idea propagated well beyond contrast detection, shaping theories of texture, motion, and object recognition, all of which now begin with a multi-scale, multi-orientation decomposition of the image that is contrast sensitivity's direct descendant.

The clinical payoff has been equally durable. Because the CSF is sensitive to disorders that leave acuity intact — early cataract, glaucoma, optic neuritis, amblyopia — it detects functional loss a letter chart can miss, and the quick, adaptive methods developed to estimate it have turned a laborious laboratory measurement into a practical test. The limitation to keep in view is that the standard CSF is measured with static, foveal, achromatic gratings; everyday vision is dynamic, peripheral, coloured, and cluttered, and sensitivity measured under laboratory conditions does not translate one-to-one to performance in natural scenes. The function is a powerful summary of spatial vision, not a complete account of seeing.

Current Directions

Three active lines extend the classical account. The first is attentional modulation: contrast sensitivity is not fixed but is raised at attended locations, and recent work has linked the effect of exogenous attention on measured contrast sensitivity to its effect on the apparent contrast of supra-threshold patterns, arguing that attention changes the early representation of contrast itself rather than only the observer's decision (Huszar, Barbot, & Carrasco, 2020). The second is the drive toward ever more efficient and informative estimation: Bayesian adaptive methods such as quick CSF and the general QUEST+ framework are being deployed to measure contrast sensitivity rapidly enough for routine clinical and even home use (Watson, 2017), with validation in low-vision populations establishing how laboratory estimates relate to functional outcomes (Stalin & Dalton, 2020). The third is the use of machine learning to connect the behavioural function to its biological substrate: deep-learning analyses of retinal imaging have begun to identify which retinal layers most constrain an individual's contrast sensitivity, pointing toward structural biomarkers of a classically psychophysical measure (Shamsi, Liu, Owsley, & Kwon, 2022).

Glossary

Centre-surround organisation.

The antagonistic receptive-field structure of retinal ganglion cells, in which a central region and a surrounding annulus have opposite signs; it suppresses broad, uniform luminance and underlies the low-frequency fall-off of the CSF.

Contrast sensitivity function.

The curve of contrast sensitivity against spatial frequency; an inverted U that peaks at medium frequencies and falls toward both coarse and fine patterns, summarising spatial vision at every scale.

Contrast threshold.

The lowest Michelson contrast at which a pattern can just be detected; contrast sensitivity is its reciprocal.

Cycles per degree.

The unit of spatial frequency: the number of light-dark cycles of a grating that fall within one degree of visual angle, so it scales with viewing distance rather than with the physical pattern.

Fourier analysis.

The decomposition of an image into sinusoidal components of different spatial frequencies, orientations, and phases; the mathematical framework that makes the grating the natural probe of spatial vision.

Grating.

A pattern of parallel light and dark stripes whose luminance varies sinusoidally; the elementary stimulus used to measure contrast sensitivity at a single spatial frequency.

Lateral inhibition.

The mutual suppression between neighbouring visual neurons that sharpens responses to edges and discounts uniform regions; a retinal mechanism contributing to the low-frequency decline of contrast sensitivity.

Michelson contrast.

A measure of a grating's contrast equal to the maximum luminance minus the minimum divided by their sum, ranging from 0 (no modulation) to 1 (full black-to-white modulation).

Multiple-channels hypothesis.

The proposal that the visual system analyses an image through many quasi-independent detectors, each tuned to a narrow band of spatial frequency; the smooth CSF is the upper envelope of these channels.

Quick CSF.

A Bayesian adaptive procedure (qCSF) that estimates the whole contrast sensitivity function in a few minutes by choosing each trial to be maximally informative about the curve's parameters.

Receptive field.

The region of visual space, and the pattern within it, to which a single visual neuron responds; its size and structure set the spatial frequencies the cell prefers.

Retinal ganglion cell.

The output neuron of the retina, whose centre-surround receptive field gives it a band-pass contrast response; the first stage at which contrast and spatial-frequency tuning appear.

Selective adaptation.

Prolonged viewing of a grating of one spatial frequency, which temporarily raises the contrast threshold for that frequency and nearby ones while sparing distant frequencies, revealing separate channels.

Spatial frequency.

How rapidly luminance varies across space, expressed in cycles per degree; low frequencies are coarse patterns, high frequencies are fine detail.

Spatial-frequency channel.

A population of neurons tuned to a narrow band of spatial frequencies that behaves as a quasi-independent detector; the building block of the multiple-channels account of the CSF.

Sub-threshold summation.

A psychophysical method in which two gratings each below threshold are combined; detection improves only when they share a channel, so the technique maps channel bandwidth.

Visual acuity.

The resolution of the finest high-contrast detail the eye can see; a single point — the high-frequency cutoff — on the contrast sensitivity function.

Key Researchers

Peter Bex

(Northeastern University) co-authored the modern synthesis on why and how contrast sensitivity should be measured, and works on natural-image and clinical contrast perception, connecting laboratory CSF measurement to real-world and low-vision visual function. ORCID

Fergus W. Campbell

(1924–1993) applied Fourier analysis to grating visibility with J. G. Robson and advanced the multiple-channels hypothesis, making the contrast sensitivity function the central tool of spatial vision. Wikipedia

Marisa Carrasco

(New York University) established how covert spatial attention alters contrast sensitivity and apparent contrast, showing that attention reaches the earliest stages of visual appearance rather than only the decision stage. ORCID

Christina Enroth-Cugell

(1919–2016) characterised the contrast sensitivity of retinal ganglion cells with J. G. Robson and introduced the X/Y cell classification, establishing that contrast and spatial-frequency tuning begin in the retina. Wikipedia

Zhong-Lin Lu

(New York University Shanghai) co-developed the quick CSF method, a Bayesian adaptive procedure that estimates the full contrast sensitivity function in minutes, transforming it into a practical clinical and research assay. ORCID

Denis G. Pelli

(New York University) co-created the QUEST adaptive method and the Pelli-Robson contrast sensitivity chart, and set out the modern principles for measuring the CSF in both laboratory and clinic. ORCID

Russell L. De Valois

(1926–2003) measured the spatial-frequency selectivity of single neurons in macaque visual cortex, giving the multiple-channels hypothesis a physiological substrate in individual frequency-tuned cells. Wikipedia

Frequently Asked Questions

What is the difference between contrast sensitivity and visual acuity?

Visual acuity measures the smallest high-contrast detail an observer can resolve, a single point at the fine end of vision. Contrast sensitivity measures how little contrast a pattern of any size needs to be seen, across all spatial frequencies. Acuity is just the high-frequency cutoff of the contrast sensitivity function, so the CSF is the more complete description of spatial vision.

What is the contrast sensitivity function (CSF)?

The CSF is a curve of contrast sensitivity plotted against spatial frequency. It has an inverted-U shape: sensitivity is highest for medium patterns, around two to five cycles per degree, and falls off for both coarse (low-frequency) and fine (high-frequency) patterns. Its overall height and the position of its peak summarise the state of spatial vision.

Why does the CSF fall off at low spatial frequencies?

Large, uniform regions are suppressed by lateral inhibition in the retina. The antagonistic centre-surround organisation of retinal ganglion cell receptive fields responds poorly to broad, slowly varying luminance, so very coarse gratings are harder to see than medium ones despite containing plenty of contrast.

What is a spatial-frequency channel?

A channel is a population of neurons tuned to a narrow band of spatial frequencies, behaving as a quasi-independent detector. The smooth CSF is the upper envelope of many such channels tiling the frequency range. Selective adaptation and sub-threshold summation experiments reveal them by showing that frequencies far apart are detected independently.

How is contrast measured?

For a grating, contrast is given by the Michelson formula: the maximum luminance minus the minimum, divided by their sum. This yields a number between 0 (no modulation) and 1 (maximum black-to-white modulation). Contrast threshold is the lowest Michelson contrast at which a pattern can be detected, and sensitivity is its reciprocal.

Why measure the whole function instead of a single threshold?

Because different disorders attack different parts of the curve. Some conditions lower sensitivity at high frequencies, mimicking an acuity loss; others reduce it at low or medium frequencies while leaving acuity normal. Measuring only acuity misses the second kind. Adaptive methods such as the quick CSF now estimate the whole function in a few minutes.

Can attention change contrast sensitivity?

Yes. Directing covert attention to a location raises contrast sensitivity there, and recent evidence indicates attention also increases the apparent contrast of supra-threshold patterns. This suggests attention modifies the early neural representation of contrast rather than only influencing the observer's decision.

What clinical conditions affect contrast sensitivity?

Contrast sensitivity loss appears in cataract, glaucoma, optic neuritis and multiple sclerosis, amblyopia, diabetic retinopathy, and normal ageing, often before visual acuity changes. Because it can reveal functional loss that a letter chart misses, contrast sensitivity testing is used to detect and monitor these conditions.

References

Robson, J. G. (1966). Spatial and temporal contrast-sensitivity functions of the visual system. Journal of the Optical Society of America, 56(8), 1141-1142. https://doi.org/10.1364/josa.56.001141

Enroth-Cugell, C., & Robson, J. G. (1966). The contrast sensitivity of retinal ganglion cells of the cat. The Journal of Physiology, 187(3), 517-552. https://doi.org/10.1113/jphysiol.1966.sp008107

Campbell, F. W., & Robson, J. G. (1968). Application of Fourier analysis to the visibility of gratings. The Journal of Physiology, 197(3), 551-566. https://doi.org/10.1113/jphysiol.1968.sp008574

Blakemore, C., & Campbell, F. W. (1969). On the existence of neurones in the human visual system selectively sensitive to the orientation and size of retinal images. The Journal of Physiology, 203(1), 237-260. https://doi.org/10.1113/jphysiol.1969.sp008862

Graham, N., & Nachmias, J. (1971). Detection of grating patterns containing two spatial frequencies: A comparison of single-channel and multiple-channels models. Vision Research, 11(3), 251-259. https://doi.org/10.1016/0042-6989(71)90189-1

Georgeson, M. A., & Sullivan, G. D. (1975). Contrast constancy: Deblurring in human vision by spatial frequency channels. The Journal of Physiology, 252(3), 627-656. https://doi.org/10.1113/jphysiol.1975.sp011162

Legge, G. E., & Foley, J. M. (1980). Contrast masking in human vision. Journal of the Optical Society of America, 70(12), 1458-1471. https://doi.org/10.1364/JOSA.70.001458

Robson, J. G., & Graham, N. (1981). Probability summation and regional variation in contrast sensitivity across the visual field. Vision Research, 21(3), 409-418. https://doi.org/10.1016/0042-6989(81)90169-3

De Valois, R. L., Albrecht, D. G., & Thorell, L. G. (1982). Spatial frequency selectivity of cells in macaque visual cortex. Vision Research, 22(5), 545-559. https://doi.org/10.1016/0042-6989(82)90113-4

Watson, A. B., & Pelli, D. G. (1983). QUEST: A Bayesian adaptive psychometric method. Perception & Psychophysics, 33(2), 113-120. https://doi.org/10.3758/bf03202828

Watson, A. B., & Ahumada, A. J. (2005). A standard model for foveal detection of spatial contrast. Journal of Vision, 5(9), 6, 717-740. https://doi.org/10.1167/5.9.6

Lesmes, L. A., Lu, Z.-L., Baek, J., & Albright, T. D. (2010). Bayesian adaptive estimation of the contrast sensitivity function: The quick CSF method. Journal of Vision, 10(3), 17, 1-21. https://doi.org/10.1167/10.3.17

Pelli, D. G., & Bex, P. (2013). Measuring contrast sensitivity. Vision Research, 90, 10-14. https://doi.org/10.1016/j.visres.2013.04.015

Watson, A. B. (2017). QUEST+: A general multidimensional Bayesian adaptive psychometric method. Journal of Vision, 17(3), 10, 1-27. https://doi.org/10.1167/17.3.10

Stalin, A., & Dalton, K. (2020). Relationship of contrast sensitivity measured using quick contrast sensitivity function with other visual functions in a low vision population. Investigative Ophthalmology & Visual Science, 61(6), 21. https://doi.org/10.1167/iovs.61.6.21

Huszar, L., Barbot, A., & Carrasco, M. (2020). Linking the effects of exogenous attention on contrast sensitivity and on apparent contrast. Journal of Vision, 20(11), 1159. https://doi.org/10.1167/jov.20.11.1159

Shamsi, F., Liu, R., Owsley, C., & Kwon, M. (2022). Identifying the retinal layers linked to human contrast sensitivity via deep learning. Investigative Ophthalmology & Visual Science, 63(2), 27. https://doi.org/10.1167/iovs.63.2.27