A dark lens can put more ultraviolet through the pupil than wearing nothing. The line that gets repeated is the pupil. Less visible light, the pupil widens, more UV gets in. The pupil does widen. It is the small term.

Bright sun makes you squint, and the lids cut the view. A tint that makes the scene comfortable holds those lids open. If the lens still passes UVA, that wider view can outweigh both the tight pupil you would have had bare-eyed and whatever UV the plastic happens to catch.

The squint is the large term

Hoover called the pupil version wrong in 1987. In a survey of more than 400 retail lenses, every one reduced UV in the crystalline lens and on the retina. The usual telling, he wrote, ignores how hard sunglass plastic already cuts the short wavelengths. He also held the view fixed — a 55° half-angle — and stopped the tally at 380 nm. Masili, Duarte, and Ventura are the ones who record those two choices.

They put the squint back in for 2024. Outdoor brightness changes how far the lids open. They paired that with a pupil model and spectra from 214 unbranded lenses: clear sky, São Paulo, 21 December (midsummer there), gaze level at the horizon, UV to 400 nm. Side light and light from behind the frame are excluded. Lens spectra were measured. Nobody parked a meter behind a living pupil. Above a ratio of 1, the glasses let more UV through the pupil than bare eyes.

Pupil area behind the lens, divided by pupil area in the sun, peaked near noon at 1.20 on the worst lens. A lens that blocked 100% of UV peaked higher, about 1.47. The darkest safe glass opened the pupil the most. Drop the lids and keep only that 1.20, and the worst lens would have looked fine.

The lids were the dose. The solid angle the eye admits, tint on versus squint intact, ran as high as 4.143 times the pupil-area term — the paper’s 314.3%. Early and late, the pupil can nudge ahead, and only slightly. Through the bright middle of the day, the open view wins.

The failures were mid-tint

Eight of the 214 crossed the ratio. None was among the darkest.

Six category 1 lenses, the lightest tint they tested, never opened the view enough to matter. Nearly clear category 0 was left out for the same reason. All 34 of the darkest still beat bare eyes: 26 in category 4, plus 8 darker than the ISO scale allows, luminous transmittance under 3%. Each of those 34 blocked at least 88% of UVA.

The eight failures sat in the middle of that darkness ladder, categories 2 and 3. Category 2 above 81% UVA block stayed protective, and category 3 above 86%. The write-up’s single line is stricter: every lens under 86% UVA block could beat bare-eye influx. UVB block did not sort them. That line is not a rack label. Shop UV400 or 100% UVA and UVB. Polarization and open temples are the other sunglasses piece.

The advice still says pupil

A 2023 note from the Pennsylvania Optometric Association has the circulating version. Darkness is not UV protection, because UV is invisible. “When you wear sunglasses, your eyes will dilate; or your pupil will widen. Therefore, if the sunglasses do not block UV rays, they can actually cause harm to your eyes.” The harm can show up. Hoover called that reason incomplete, and this run left it small. A larger pupil behind a lens that passes no UV does not raise the tally. Lids held open, with UVA still getting through, can.

The useful half of that note is the American Optometric Association’s bar: 99–100% of UVA and UVB.

Before you trust the tint

1. Read today’s peak on the homepage. WHO’s protect line for eyes and skin is UV index 3. The index will not say whether a lens blocks UVA. Shade, clothing, and sunscreen for the same outing are in the Guide.
2. Read the UV line before the darkness. UV400 or 100% UVA/UVB. The mid-80s figure is a failure line inside one calculation, not a pass mark.
3. Comfort means the squint is off. You will not get the lid closure back while the pair stays on. The filter has to be in the lens.
4. Close the sides when the ground is bright. This model ignored UV bouncing in from snow, water, and sand. Wrap and a brim cover that path.
5. Keep a labeled pair. The other 206 lenses did not exceed bare eyes. Eight bad ones is not a reason to skip UV400.

One clear-sky year, one city, horizon gaze, no clouds. Hoover, squint locked, found no lens that increased dose. This run found eight, and only with the lids free. Same answer on the pupil. A different answer on how often a pair loses to bare eyes.

Dose math for planning. Not a diagnosis.

Sources

Pupil dilation is the wrong driver, and the 400-plus retail survey: Hoover, Applied Optics, 1987. The 55° field and the 380 nm cutoff are Masili, Duarte, and Ventura’s account of that setup. Squint versus pupil, 214 lenses, the 1.20 and 1.47 ratios, the 4.143 / 314.3% field-of-view term, eight failures, the 81% and 86% UVA lines, and the 88% floor on the darkest 34: Masili, Duarte, and Ventura, Scientific Reports, 2024. Pupil wording and the 99–100% bar: Pennsylvania Optometric Association and the American Optometric Association. Protect from UV index 3: WHO.