Look at the picture below.
Two things change as you move across it. Move left to right and the stripes get finer — more cycles in the same width. Move bottom to top and the stripes get fainter — the light bands less light, the dark bands less dark, until they disappear into the gray. The bottom is dense; the top is featureless.
But here's the trick. If you scan upward in any single column, the moment at which the stripes vanish is not at the same height in every column. In the middle of the chart you can see them most of the way up. At the very left and the very right, they fade out far sooner. The visible region traces out a curve that bulges upward in the middle and drops at both ends.
That curve is your contrast sensitivity function. The chart is doing a measurement on you, by eye, that ordinarily needs a calibrated instrument and a few hundred trials. It is one of the most quietly remarkable images in vision science, and the rest of this post is about how to read it.
Trace the upper edge of where you can see the stripes. That edge is your contrast sensitivity function.
What the chart is doing
The image you just looked at is called a Campbell-Robson chart. It was introduced by Fergus Campbell and John Robson in their 1968 paper in the Journal of Physiology, which laid out the spatial-frequency framework that almost all modern vision science still uses (Campbell & Robson, 1968). The figure is so neat and so explanatory that the field absorbed it whole.
Two axes, encoded in the image itself.
The horizontal axis is spatial frequency — how fine the pattern is. Spatial frequency is usually measured in cycles per degree of visual angle (cpd): how many light-dark stripe pairs fit inside one degree of your visual field. A degree of visual angle is roughly the width of your thumb at arm's length. The far left of the chart shows around one stripe pair per thumb-width — coarse. The far right shows dozens — fine. The progression is logarithmic, so that equal horizontal distances correspond to equal ratios of frequency. That matches how the visual system itself responds to spatial scale: in approximately log steps, not linear ones.
The vertical axis is contrast — how different the dark stripes are from the light ones. Contrast at the very bottom is maximum (black against white). At the top it is near zero (almost the same shade of gray either way). In the original 1968 Campbell-Robson figure, contrast was plotted increasing downward on a log scale; we follow that convention here. So the bottom row contains the full-contrast stripes; the top row contains stripes so faint they merge into the surround.
The pattern itself is a single continuous sine-wave grating: light bands fading smoothly into dark bands rather than hard edges. Sine-wave gratings are the standard stimulus in contrast-sensitivity research because they are the cleanest possible input — they contain only one spatial frequency, not the broadband mixture of frequencies you would get from sharp-edged stripes or from letters. Each column of the chart contains stripes at one specific spatial frequency, repeated all the way up and down. Crucially: the physical contrast of the stripes is the same up and down within any single horizontal row. A horizontal slice through the chart has constant physical contrast — what changes is only the stripe pitch.
That is the part most people miss on the first look. The contrast at any single height is the same across the whole image. The visibility is not.
Reading your own curve from it
Here is the operation. Pick any column — say, somewhere on the left. Start at the bottom (where contrast is full, and the stripes are loud) and let your eye scan slowly upward. At some height, the stripes will fade out into the gray surround. Mark that height in your head.
Now do the same for a column in the middle of the chart, around the peak of the inverted-U region.
Then a column at the far right, where the stripes are fine.
For most healthy adults the height at which the stripes vanish is highest in the middle and lower at both ends. Your eye traces an arch — a curve — across the top edge of the visible region. That arch is your contrast sensitivity function (CSF). The vertical position of the arch at each column is your contrast threshold at that spatial frequency; the inverse of threshold is sensitivity. Plot height versus column, log-log, and you have the same inverted-U we sketched out in the primer on what contrast sensitivity actually measures.
A normal young adult's curve peaks around 3 to 6 cpd. That is the band you use to recognise a face across a room, read low-contrast print, see a curb against pavement at dusk. It is the part of the visual world most people find the most rewarding to be good at. It is also the part that quietly drops first in conditions like cataract, glaucoma, multiple sclerosis, post-concussion vision change, and ordinary aging — sometimes well before high-contrast acuity (the "20/20" number from your last eye exam) shows any change at all.
A couple of practical caveats before you try to read off a clinical number.
Viewing matters. The horizontal axis is in cycles per degree of visual angle, which means the exact cpd value any column corresponds to depends on how big the image is on your screen and how far away your eye is. Sit at roughly arm's length from a laptop and you are looking at a wide range of spatial frequencies — from very coarse (well under 1 cpd) to very fine (well over 20 cpd). Move closer, and every column shifts to a lower cpd; move farther, and they all shift up. Print the chart on letter paper held at 50 cm, and you will see the same shape at a calibrated scale.
Your screen matters. A dim or glossy display, a non-standard gamma, a brightness setting that is fighting an over-lit room — all of these stretch or compress the visible region, sometimes badly. If you can dim the lights and put your screen brightness near maximum on a calibrated monitor, the chart shape will read closer to truth.
You are one observer, not a hundred. Eyeballing the chart is a beautifully informative qualitative demonstration. It is not a measurement with error bars.
Why this is, gently, a magic trick
Here is the part worth sitting with for a moment.
The columns of this chart are physically identical in contrast top to bottom — by row. We baked that in. The image is generated mathematically, and every horizontal row of pixels has been given the same Michelson contrast (the standard measure of how different the light and dark portions of a sinusoid are). If you took a photometer to the screen and measured the light-dark difference at any height, the value would not depend on whether the column was at the left edge or the right edge.
And yet the visibility of the stripes — what your visual system does with that physical contrast — is wildly different across the chart. Stripes in the middle of the chart are vividly present halfway up the image. Stripes at the far edges have already vanished into gray. The reason cannot be in the image, because the image is uniform across rows. The reason has to be in you.
That is what Campbell and Robson set out to show. Their argument, made formally with Fourier analysis, was that the visual system applies a frequency-dependent filter to every image it sees. Low spatial frequencies are partly suppressed by lateral inhibition between neighbouring retinal cells; high spatial frequencies are limited by the eye's optics and by the spacing of cone photoreceptors. In the middle band — around the human peak of about 3 to 6 cpd — the filter is at its most generous. The Campbell-Robson chart turns that filter into a picture. It is the only common visual stimulus we know of where the shape of the perceptual transformation is plotted directly, in the image, as visible geometry.
This is, in the literal sense, what got a generation of vision scientists hooked. You can read a hundred pages on the CSF and not feel it the way you feel it after one minute with this chart.
What our test does that the chart can't
Eyeballing the Campbell-Robson chart is a qualitative demonstration. Measuring your CSF properly is a quantitative problem, and it is harder than it looks. A single observation of "I can see them up to about here" carries a lot of noise: where you decided to stop, how long you stared, whether you blinked, what the surround luminance was when you started.
Clinical contrast sensitivity tests handle this by repeating the question dozens or hundreds of times under controlled stimulus parameters and converging on a threshold statistically. The classic adaptive approach is a transformed up-down staircase — present a stimulus, ask the observer to identify it, make it harder on each correct answer and easier on each incorrect one, and average over the last several reversals (Levitt, 1971). The more efficient modern approach is QUEST, a Bayesian procedure that places each new trial where it will most reduce uncertainty about the threshold given everything the observer has done so far (Watson & Pelli, 1983).
Our test runs an adaptive staircase at several spatial frequencies and gives you back the curve. It takes about five minutes, calibrates against your screen and your viewing distance, and produces a numeric log contrast sensitivity at each frequency — comparable, with caveats, to the kinds of measurements that clinical chart tests like the Pelli-Robson letter chart were designed to produce (Pelli, Robson & Wilkins, 1988). If you've found the Campbell-Robson chart visually compelling and want a real number to look at, that is the next step. For more on how the calibration works, see the methodology page. For a tour of the major clinical CS tests, see the Pelli-Robson vs FACT vs qCSF comparison.
A reduced CSF is a screening signal, not a diagnosis of anything. But it is worth knowing your baseline so that you can notice if it changes.
Want a real measurement?
Take the test. Free, open methodology, results stay on your device by default. Five minutes. Then come back and look at the chart again — you will see it differently.
References
- Campbell, F. W., & Robson, J. G. (1968). Application of Fourier analysis to the visibility of gratings. The Journal of Physiology, 197(3), 551–566. The foundational paper. The Campbell-Robson chart appears as Figure 1, and the spatial-frequency-channel framework laid out in the rest of the paper is still the framework in use today.
- Pelli, D. G., Robson, J. G., & Wilkins, A. J. (1988). The design of a new letter chart for measuring contrast sensitivity. Clinical Vision Sciences, 2, 187–199. The Pelli-Robson letter chart, the most widely used clinical contrast-sensitivity test. Reports a test-retest repeatability of about ±0.15 log units and a smallest clinically meaningful change of about ±0.30 log units — useful numbers to keep in mind when interpreting any single CS measurement.
- Watson, A. B., & Pelli, D. G. (1983). QUEST: a Bayesian adaptive psychometric method. Perception & Psychophysics, 33(2), 113–120. The Bayesian threshold-estimation procedure underlying most modern efficient contrast-sensitivity tests.
- Levitt, H. (1971). Transformed up-down methods in psychoacoustics. The Journal of the Acoustical Society of America, 49(2B), 467–477. The classic paper on adaptive staircase procedures (originally psychoacoustic; the methods are general and were imported into vision research wholesale).