You can have 20/20 vision and still struggle to drive home at dusk. You can pass the eye chart in your optometrist's office and miss the edge of a curb in the fog. You can read the smallest line on a wall ten feet away and find a friend's face in a dim restaurant strangely hard to recognize. People who have had a concussion describe something stranger still: the letters on the chart are crisp, but the world has somehow flattened, and they cannot say why.
The puzzle is that the eye chart isn't testing what those situations require. It tests one thing — your ability to resolve very fine, very high-contrast detail at a fixed distance — and gives you a single number for it. Real vision is bigger than that. The measurement that fills in the rest of the picture is called contrast sensitivity, and the rest of this post is about what it is, what it measures, and why it matters that we measure both.
The first answer most clinics give: visual acuity
The Snellen chart — those rows of black letters on white, shrinking as you go down — has been the standard eye-exam tool since Herman Snellen designed it in 1862. It is fast, cheap, requires only a wall and a printed chart, and produces a clean number anyone can repeat: 20/20, 20/40, 20/15. The numerator is the testing distance in feet; the denominator is the distance at which a person with "normal" vision could read the smallest line you managed.
Visual acuity is a real measurement of a real ability. It captures how finely your visual system can resolve detail at high contrast — when the letter is a hard black against a hard white. That is genuinely useful information. If your acuity is bad, your glasses prescription is probably wrong, or something is going on with the optics at the front of your eye, or a part of your retina or visual pathway can no longer relay fine detail. Catching those things early is the whole point of an eye exam, and the Snellen chart catches a lot of them.
Acuity became standard because of three useful properties: it is objective (the letter is either named correctly or not), portable (a printed chart works anywhere), and comparable (everyone's 20/20 means roughly the same thing). For more than 150 years, "20/20" has been the shorthand for "your vision is fine."
The problem is the second half of that sentence. The chart can certify that one specific capability is intact. It cannot certify that vision, as a whole, is working well.
What the eye chart misses
Almost nothing you look at in a day is a hard black letter on hard white paper at twenty feet. Most of what your visual system handles is somewhere in the middle: a face in a dim room, a road sign in fog, a curb against the sidewalk, a step against the rug, a colleague's expression across the table at a coffee shop. The relevant contrast — the luminance difference between the thing and its background — is rarely the 99% contrast of ink on paper. It might be 5%, or 1%, or 0.5%.
A second thing the chart misses is that vision is a curve, not a single point. Things in the world have different sizes and different contrasts, and the visual system handles them differently depending on both. A coarse pattern at low contrast (a foggy outline of a tree) and a fine pattern at low contrast (the faint texture of newsprint) are very different signals; the brain extracts them through different channels and with different efficiency. Acuity samples one extreme of one of those axes — the highest-contrast, finest-detail corner. Almost everything else is uncharted on the eye chart.
If you imagine the space of visual stimuli as a 2D grid — detail on one axis (coarse to fine) and contrast on the other (high to low) — the Snellen chart is one dot in the corner. Contrast sensitivity is the map of the rest of the room.
This is not a hypothetical gap. There are conditions — early glaucoma, early cataract, multiple sclerosis, some forms of post-concussion vision change, diabetic retinopathy — where contrast sensitivity is measurably reduced before acuity drops at all (see the StatPearls overview of contrast sensitivity for a clinical summary). The patient passes the eye chart and goes home with a clean bill of visual health, and the thing affecting their daily vision goes unmeasured.
Contrast sensitivity, defined precisely
Here is the technical definition, broken into pieces.
Contrast is the difference in luminance between adjacent regions of an image. For a striped pattern — say, a series of light and dark bars — the standard measure is Michelson contrast, which is the ratio of the difference between the brightest and darkest bars to their sum. Two strips that are identical have a Michelson contrast of 0. A black bar next to a white bar at twice the average brightness has a Michelson contrast of 1. Most things in the world are somewhere in between: a faded road sign might be 10% contrast against the sky; a face in evening light might be 5% against its background.
Contrast threshold is the lowest contrast at which you can still reliably see a given pattern. If a pattern of stripes is barely visible to you at 2% contrast and invisible at 1%, your threshold for that pattern is somewhere around 1.5%.
Contrast sensitivity is just the inverse of contrast threshold. Sensitivity = 1 / threshold. If your threshold is 1%, your sensitivity is 100. If your threshold is 0.5%, your sensitivity is 200 — meaning you can see a pattern at half the contrast someone else needs. Higher sensitivity, better vision.
Because sensitivity ranges across orders of magnitude (typical values run from about 1 to about 500 across different pattern sizes), the conventional unit is log contrast sensitivity — log base 10 of the sensitivity number. So a sensitivity of 100 is a log CS of 2.0; a sensitivity of 200 is a log CS of about 2.3. The log scale keeps the numbers manageable and lines up with how the visual system itself encodes differences (in roughly logarithmic steps, not linear ones).
Now the key claim. Visual acuity — your "20/20" — sits near one end of this measurement. Specifically, it sits near the high-spatial-frequency cutoff, where the pattern is so fine that you need essentially perfect optics and retinal sampling to resolve it. That single point is useful, but it tells you very little about how you handle the medium-detail, medium-contrast world that makes up most of vision.
Contrast sensitivity, measured properly, is a curve across a range of pattern sizes. The curve has a name: the contrast sensitivity function.
The contrast sensitivity function (CSF)
The contrast sensitivity function — CSF — plots how sensitive you are to patterns of different sizes. The x-axis is spatial frequency: how many cycles of pattern fit into one degree of your visual field, written in cycles per degree or cpd. Coarse patterns (a few thick bars) are at the low-frequency end; fine patterns (many thin bars) are at the high-frequency end. The y-axis is sensitivity — how faint a pattern at that size you can still see.
For a healthy young adult, the curve looks roughly like an inverted U, plotted on log axes:
The curve has three regions worth knowing:
- Low spatial frequency (left side, coarse patterns). Sensitivity is reduced here, in part because the visual system actively suppresses very-low-frequency information through lateral inhibition between neighbouring cells in the retina. This is what lets you see edges sharply; the cost is that very smooth, very gradual changes in brightness are not where your system shines.
- Peak (around 3 to 6 cpd). The most sensitive region. Most of the work of recognising faces, reading body language, and distinguishing one object from another at conversational distance happens in this band.
- High spatial frequency (right side, fine detail). Sensitivity falls steeply. The right-hand tail of the curve eventually crosses the x-axis at the acuity cutoff — about 50 to 60 cpd in a healthy eye, which is roughly where 20/20 lives.
So 20/20 is not "vision." It is one point on the right-hand tail of this curve.
The CSF as a curve was first established by Fergus Campbell and John Robson in their 1968 paper in the Journal of Physiology, which laid out the spatial-frequency framework that the field still uses. Their core observation was that the visual system behaves as if it contains separate channels tuned to different pattern sizes — and that measuring sensitivity across those channels reveals far more about visual function than any single test of acuity could (Campbell & Robson, 1968).
Note: a CSF measurement is a screening signal of overall visual function — useful for tracking change over time and for prompting a conversation with a clinician, not a diagnosis of any specific condition.
What this lets you see that 20/20 doesn't
Different conditions affect different parts of the curve. The mapping isn't perfectly specific — many conditions overlap — but the patterns are informative.
- Cataract generally lowers the whole curve, with the strongest effect on mid-to-high frequencies, because intraocular scatter reduces effective contrast across the board.
- Glaucoma preferentially affects the magnocellular pathway, and mid-frequency contrast sensitivity losses can appear before visual field defects show up on standard perimetry.
- Multiple sclerosis and optic neuritis are associated with broad CSF reductions, often centred in the mid-to-high range. Low-contrast acuity is being added to MS clinical-trial outcome measures for this reason.
- Diabetic retinopathy is associated with early CS losses that can precede visible retinopathy.
- Concussion and mild traumatic brain injury are associated with reductions in mid-frequency contrast sensitivity in published studies of first- and second-order CSFs.
- Normal aging reduces high-frequency sensitivity gradually — roughly 10% per decade after age 20.
These are associations and patterns reported in the literature, not diagnostic shortcuts. None of them are visible on a Snellen chart until they become severe. All of them produce a measurable change in CSF that an attentive screening — like the test we built — can register early enough to be worth a conversation with a clinician.
Future posts here will go into each of these conditions in detail. If one is on your mind, a link to that post will appear at /blog/[condition]-and-contrast-sensitivity as it's published.
What contrast sensitivity doesn't do
Contrast sensitivity, however measured, is a screening signal. It is not a diagnostic test.
A single CSF measurement is sensitive to a long list of incidental things: how rested you are, whether you're wearing your current glasses prescription, the brightness and contrast of the screen you're testing on, the room lighting, whether you've recently been on a long drive, and how the test was calibrated. Test-retest variation is real even with the most carefully validated instruments — the Pelli-Robson chart, for instance, has a test-retest repeatability of about ±0.15 log units; the smallest clinically meaningful change is usually taken to be about ±0.30 log units (Pelli, Robson & Wilkins, 1988).
Online tests have additional sources of variance: display gamma, pixel pitch, viewing distance, ambient light, and the user's own setup all influence the result. A well-built online test takes that seriously and tries to control for it (we do — see /methodology), but no remote screening can be as tightly controlled as a clinical instrument with a calibrated luminance standard.
What this means in practice:
- A single CSF result is a snapshot, not a verdict.
- A change over multiple sessions on the same setup is more meaningful than any single absolute number.
- A reduced result is a reason to make an eye-exam appointment, not a diagnosis of anything.
- A normal result does not rule out conditions that don't primarily affect contrast — get a real eye exam if you have new visual symptoms.
The honest framing: contrast sensitivity is a more thorough functional measurement than acuity, and a useful complement to it. It is also a noisier one, especially in unsupervised settings, and it should be read as a signal among many.
Where to go from here
If you want to see what your own CSF looks like, you can take a free test in your browser. It runs an adaptive psychophysical procedure across several spatial frequencies, calibrates against your screen, and plots a curve you can read and share with a clinician. Results stay on your device by default.
If you'd rather keep reading first, the next posts in this series cover the major clinical tests of contrast sensitivity (Pelli-Robson, FACT, qCSF), what screen calibration has to do with remote testing, and the condition-specific patterns above in more detail.
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 establishing the contrast sensitivity function and the spatial-frequency channel framework.
- 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 chart paper, still the most widely used clinical letter-based CS test.
- Mäntyjärvi, M., & Laitinen, T. (2001). Normal values for the Pelli-Robson contrast sensitivity test. Journal of Cataract and Refractive Surgery, 27(2), 261–266. Source of the age-stratified normative values quoted in clinical practice (e.g., monocular log CS ~1.84 at age 20–39, ~1.68 at age 60+).
- 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. The qCSF method — a Bayesian adaptive procedure that estimates the entire contrast sensitivity curve in a few minutes.
- Watson, A. B., & Pelli, D. G. (1983). QUEST: a Bayesian adaptive psychometric method. Perception & Psychophysics, 33(2), 113–120. The foundational efficient-threshold paper underlying modern adaptive contrast-sensitivity testing.