You leave the office a little late. The sky is still bright in the west, the eastern half has gone slate, and the streetlights have just clicked on but haven't yet caught up. For about thirty minutes the world sits in a particular kind of half-light. A jogger crosses the road in a dark hoodie and you see them later than you wanted to. A car the other way has its headlights on, and after it passes there are a few seconds during which the road looks washed out. The reflective sign you usually read at a glance takes a beat longer to resolve.
This is the part of the day when your visual system is doing the most adapting and getting the least help from the road's design — and consistently one of the highest-risk windows for driving, especially for pedestrian collisions. The reason is not that your eyes have suddenly become bad. You are operating in a regime — vision scientists call it mesopic — that lies between two sets of rules, and the roads, the cars, and the standard eye chart are designed around the rules at either end. The rest of this post is about what is happening in that window, why some people lose more contrast in it than others, and what you can usefully do.
A 30-minute window between two regimes
The drop in light between sunset and full dark is fast — roughly 2 to 3 orders of magnitude over half an hour, depending on latitude and weather. Crash statistics from national road-safety agencies have for decades pointed at this window as disproportionately dangerous, especially for pedestrians: the relative risk per vehicle-mile climbs steeply as ambient light falls and stays elevated until headlight illumination dominates the scene. Traffic is still close to its evening peak; pedestrians are still numerous; the road is no longer reliably illuminated by the sky and not yet by anything else.
The reason the same driver who handled the same road at 5 p.m. struggles with it at 7 p.m. is, in part, that the visual system itself works differently in that range. Bright daylight is the photopic regime: cones run the show, colour vision is intact, the pupil is small, and the contrast sensitivity function sits at its highest point. Moonless dark is scotopic: rods carry the load, there is no colour, and the curve drops by an order of magnitude. Mesopic is everything between — both rods and cones contribute, the system is in transition, and the curve sits lower than daylight, shifted toward coarser detail, and is more sensitive to individual differences in optics and physiology. The full mechanism story — pupil dilation, photon noise, rod-cone shift — is in our low-light contrast post; this post stays on the road.
What makes dusk specifically hard is not just that you are mesopic, but that you are becoming mesopic. Cone sensitivity has been falling for a few minutes; rod sensitivity will take another fifteen to twenty to come fully online. You are operating on whichever pieces of the system happen to be available right now, and those pieces are themselves changing minute to minute as you drive.
What goes wrong on the road
Three failure modes account for most of the dusk-specific risk, and each maps onto a particular gap between how vision works in that window and how the road is designed.
Pedestrians in dark clothing. A jogger in black on a darkening road surface is one of the lowest-contrast targets your driving day will produce. In photopic daylight a dark silhouette has the sky and the road behind it to provide a contrast frame, and your CSF is at its peak. In full dark the headlights' retroreflective return from the road surface makes the silhouette pop. In the dusk window neither is true: the sky is grey rather than bright, the road is dim rather than headlamp-bright, and the silhouette is a low-contrast shape on a low-contrast background — exactly the kind of target a mesopic CSF handles worst. This is the worst-case driving target in the worst-case lighting regime, and it explains a great deal of the pedestrian-fatality skew toward this hour.
Oncoming headlights and glare recovery. The eyes that are doing your driving at dusk are already mid-adaptation. When an oncoming car's headlights cross your field they push a large amount of light into a partly-dark-adapted eye, and the time it takes your visual system to recover its low-contrast sensitivity after the source has passed is longer than it would be in either daylight or fully dark conditions. The pupils have to re-dilate, the photoreceptors have to re-adapt, and the cortical contrast-gain control has to re-settle — all on top of an adaptation state that was already moving. Owsley and colleagues' long line of work on aging vision and driving documents the role of glare recovery in real-world driving performance, with the strongest effects in patients whose ocular optics are already imperfect (Owsley, Stalvey, Wells, Sloane & McGwin, 2001).
Road signs caught between regimes. Road signs are engineered for two specific lighting scenarios: daylight, where the sign reflects ambient sunlight, and headlight illumination at night, where retroreflective material returns the headlamp light to the driver at much higher intensity than the surrounding scene. The dusk window gets neither: ambient sunlight has dropped too far for the daytime case, and headlights are not yet bright enough relative to the residual sky to drive the retroreflective case. The same sign you read in a glance at noon takes measurably longer at twilight.
Underneath these three is a fourth: familiarity. People often drive their dusk routes from memory — the same commute, the same lane changes, the same lit and unlit stretches. Familiarity makes the visual demands feel lower than they are. The pedestrian who steps off the curb where pedestrians have not stepped off before is exactly the case where memory does not help.
What changes between daylight and dusk in your own eyes
Even in a healthy young adult, the contrast sensitivity curve at mesopic luminance is not the daytime curve. The peak gain is lower, the curve is shifted slightly toward coarser spatial detail, and the high-frequency arm — the part that fine-detail tasks live on — falls away faster. That is the universal picture.
The catch is that the amount by which any given person's curve drops between daylight and dusk varies far more than the daytime curves themselves do. Two drivers with indistinguishable Snellen acuity and Pelli-Robson contrast sensitivity in office lighting can be very different drivers at dusk, because the modifiers of mesopic vision — pupil size and aberration at full dilation, intraocular scatter, tear-film stability, retinal sensitivity — vary across people more than the photopic equivalents. Joshi and Sivaprasad's 2023 review summarises the pattern: dropping the luminance level shifts every age band's curve downward, with the steepest impact on the high spatial frequencies and a meaningful inter-individual spread not predictable from a single daylight measurement (Joshi & Sivaprasad, 2023).
The honest punchline: you do not know how badly your contrast sensitivity drops at dusk until something measures it. A daytime test gives you a strong hint — drivers with reduced photopic contrast sensitivity tend to drop further in dim conditions, not less — but it does not give you the dusk number directly.
Who carries extra risk in this window
Dusk amplifies whatever else is going on. Drivers whose contrast sensitivity is already a little reduced in photopic conditions tend to be more than a little reduced in mesopic ones. The groups for whom this matters most are the ones whose CSF was already moving — usually quietly, often without an obvious symptom story in the daytime.
Older drivers. Contrast sensitivity declines gradually with age — roughly 10% per decade after age 20 (Owsley, Sekuler & Siemsen, 1983; Mäntyjärvi & Laitinen, 2001) — and the maximum pupil dilation also diminishes. An older eye in a dim regime gets less retinal illuminance for any given ambient light level and has less contrast sensitivity to work with. The drop from daylight to dusk is steeper for an older driver than a younger one, even if both are clinically healthy.
Early cataract. Intraocular scatter is the single most contrast-destroying thing the front of the eye can develop, and it does its worst work in dim conditions where the small daytime pupil is no longer masking the scattering periphery of the lens. The pattern of "20/20 acuity but headlight glare is harsher than it used to be and the dusk drive feels different" is the canonical early-cataract presentation — covered in the cataract and night-driving piece.
Glaucoma. Glaucomatous damage to retinal ganglion cells reduces contrast sensitivity, often before standard perimetry detects a visual-field defect. Magnocellular-pathway signal — which carries the low-spatial-frequency, motion-relevant information you depend on when scanning a darkening road — is preferentially affected.
Multiple sclerosis and post-optic-neuritis vision. Contrast sensitivity changes are well documented after optic neuritis and in MS more generally, and they can persist after acuity has recovered. The drop into mesopic conditions can re-expose deficits that are largely invisible in daylight (see the MS and contrast post).
Post-concussion vision. The mid-band contrast sensitivity changes that follow some concussions tend to be more symptomatic in low-contrast, low-light tasks than in bright high-contrast ones. A driver in concussion recovery often reports that the dusk drive is the hardest part of the day.
If you are in one of these groups, dusk driving deserves a different level of attention than the daytime drive does — not because the eye exam said something specific, but because the regime amplifies whatever the eye exam might not have caught.
What to do, practically
Some of this is sensible-driving boilerplate. Some of it is specific to the mesopic regime and worth doing on purpose.
- Get an eye exam every one to two years if you drive at dusk regularly, more often after 60. Mention dusk driving specifically — it gives the clinician a reason to look at parts of vision the standard battery does not always reach.
- Track your contrast sensitivity. Take the test once to establish a baseline, then every six to twelve months on the same device. A trend is harder to misread than a single result.
- Increase following distance and slow down. You are not the same driver you were at 4 p.m.; behave like a driver whose reaction time is a beat slower. The marginal cost of 5 mph is small; the benefit in stopping distance against a low-contrast target is large.
- Scan the shoulders. Pedestrians and cyclists in dark clothing live in the road's lowest-contrast region. Train the habit of glancing at the road's edge in the dusk window.
- Do not lean on familiar-route memory. This is the hour when the difference between the road you remember and the road you can actually see is largest.
- Ask your optometrist about anti-reflective coatings on glasses. A meaningful share of dusk discomfort comes from internal reflections off the back surface of spectacle lenses when there is glare in the scene. Anti-reflective coatings reduce that reflection and are inexpensive relative to the symptom relief.
What our test can — and cannot — tell you
Note. A contrast sensitivity test is a screening signal of overall visual function, not a diagnosis of any condition. A lower-than-typical result is a reason to take a better question to your eye doctor, not a verdict. Many of the conditions referenced above can be diagnosed and addressed by a routine eye exam — but only one if the conversation goes beyond the standard chart.
Our test is designed for normally-lit indoor conditions. It measures photopic contrast sensitivity — the daytime curve, not the dusk one. That is an honest limit, and naming it is a better choice than pretending we measure something we don't. There is no widely available consumer-grade test that runs at mesopic luminance on consumer screens.
The reason a photopic test is still useful for thinking about dusk driving is that a reduced photopic contrast sensitivity reliably predicts an even more reduced mesopic one. The two curves are not the same, but they move together: a reduced photopic result is a strong signal that the mesopic result is also reduced — and probably more so. For a dedicated mesopic test, an optometrist's office is the right next stop; some clinical instruments (the CSV-1000HG and a handful of mesopic-with-glare protocols) run the test at controlled low luminance.
Take the test
Take the test now. Three minutes, a normally-lit room, a saved result. If your photopic contrast sensitivity is below typical for your age, consider an eye exam before your next dusk drive — and mention dusk driving by name. Calibration and adaptive-procedure details for our test live on the methodology page; the broader background on what contrast sensitivity measures lives in the primer post.
References
- Owsley, C., Sekuler, R., & Siemsen, D. (1983). Contrast sensitivity throughout adulthood. Vision Research, 23(7), 689–699. The reference dataset for how contrast sensitivity declines across the adult lifespan, with the steepest age effects in the high spatial frequencies relevant to driving.
- Ball, K., Owsley, C., Sloane, M. E., Roenker, D. L., & Bruni, J. R. (1993). Visual attention problems as a predictor of vehicle crashes in older drivers. Investigative Ophthalmology & Visual Science, 34(11), 3110–3123. The Useful Field of View work linking visual processing deficits — which include contrast-relevant components — to subsequent crash involvement in older adults.
- Owsley, C., Stalvey, B. T., Wells, J., Sloane, M. E., & McGwin, G., Jr. (2001). Visual risk factors for crash involvement in older drivers with cataract. Archives of Ophthalmology, 119(6), 881–887. Older drivers with cataract and reduced contrast sensitivity had elevated crash risk; standard visual acuity was a weaker predictor on its own.
- Owsley, C., McGwin, G., Jr., Sloane, M., Wells, J., Stalvey, B. T., & Gauthreaux, S. (2002). Impact of cataract surgery on motor vehicle crash involvement by older adults. JAMA, 288(7), 841–849. Cataract surgery was associated with a meaningful reduction in subsequent motor vehicle crashes compared with deferred surgery.
- Mäntyjärvi, M., & Laitinen, T. (2001). Normal values for the Pelli-Robson contrast sensitivity test. Journal of Cataract and Refractive Surgery, 27(2), 261–266. Age-stratified normative Pelli-Robson values; the source for "normal" contrast sensitivity by decade used in clinical practice.
- Joshi, M. R., & Sivaprasad, S. (2023). Spatial contrast sensitivity function and its neurophysiological bases. Progress in Retinal and Eye Research, PMC10527080. Review of how the CSF depends on retinal and cortical processing, and how luminance regime and disease shift the curve.