The science
Training the brain to de-blur
Presbyopia is an optical problem: the lens stops focusing up close. Its effect on reading also depends on how efficiently the visual cortex processes a blurred, low-contrast image. Kestrel trains that processing.
1 · What goes wrong with age
After about 40, the eye's lens loses its ability to change shape (accommodate). Up close, fine detail (high spatial frequencies) is blurred before it reaches the retina. A blurred image has lower contrast, and lower contrast drives weaker and slower responses in visual cortex. The result is smaller letters you can't resolve, slower reading, and fatigue.
2 · Perceptual learning
Repeated practice on a demanding visual task, at the limit of your ability, can permanently improve that task and sometimes related ones. Researchers first showed this in amblyopia (lazy eye) and then applied it to presbyopia. The hypothesis: if neurons respond with more gain and less noise, and integrate information faster, the brain can extract a usable signal from a blurry input.
3 · The exact training paradigm
Kestrel implements the protocol described in the Methods of Polat et al. (2012) and Lev et al. (2014):
- Stimuli: Gabor patches: striped gratings in a soft Gaussian window (σ = one wavelength). They match the receptive fields of neurons in primary visual cortex. Spatial frequencies run 2–8 cycles/degree; orientations 0°, 45°, 90° and 135°.
- Task: two-interval forced choice. Two flashes, each marked by a tone; the low-contrast target is in one. You pick which.
- Lateral masking (LM): the target sits between two collinear high-contrast flankers (40% contrast), 2–4 wavelengths away. At these distances, collinear flankers normally facilitate detection through horizontal connections in V1 (Polat & Sagi, 1993–94).
- Backward masking (BM-T and BM-LM): a second flanker pair appears 60–240 ms after the target. A mask that arrives quickly enough cancels the flankers' facilitation. Training pushes processing to finish before the mask, so the mask delay shrinks over the program.
- Each session: four blocks (target alone, LM, BM-T, BM-LM), each with an adaptive 3-down/1-up staircase that converges on ≈79% correct.
- Between sessions: difficulty changes one parameter at a time based on the previous session. Kestrel uses a ladder of mask delay (240 → 60 ms) × spatial frequency (2 → 8 c/deg), with presentation time 90 → 60 ms. The masking index, log(BM-LM threshold ÷ LM threshold), decides when to advance.
- Viewing: 40 cm, both eyes, no reading glasses, dim room, maximum brightness. 30–40 sessions on separate days.
200 ms
0.3–1.3 s
60–90 ms
60–240 ms
60–90 ms
4 · Making it precise on a home screen
- Size: you match an on-screen outline to a bank card (85.6 mm) once. Kestrel then computes pixels per degree at 40 cm and draws every pattern at its true angular size. It only uses spatial frequencies your screen can render with at least 4 pixels per cycle.
- Contrast: luminance is computed in linear light and converted with the sRGB curve. Ordinary 8-bit screens can't show contrasts below ~0.4%, so Kestrel uses noisy-bit dithering (Allard & Faubert, 2008) to render sub-step contrasts accurately.
- Timing: flashes are locked to your display's refresh rate, measured at calibration. If the browser drops a frame during a flash, the trial is discarded and repeated.
5 · Evidence
Polat et al., 2012, Scientific Reports. 30 presbyopes (age 51 ± 4) completed 37 ± 11 sessions over 3 months. Uncorrected near acuity improved in every participant, from 2.44 to 1.56 arc-min on average (≈1.6×, about 2 ETDRS lines). The authors describe this as equivalent to being ~8.6 years younger. Reading speed at the smallest print rose ~17 words/min, and contrast sensitivity improved 19–34%. Accommodation, pupil size and depth of focus did not change, so the gains were neural. Three untrained controls showed no change.
Lev et al., 2014, Scientific Reports. 14 young adults trained with the mobile version (24 sessions) against 19 untrained controls. Training increased processing speed and transferred to untrained tasks: crowded acuity, and far as well as near acuity.
Yehezkel et al., 2016, Scientific Reports. 119 adults aged 40–63. Near-acuity gains were proportional to how impaired vision was at the start. People with worse near vision gained more; those already near 20/20 gained little.
Sterkin et al., 2018, Vision Research. Presbyopic pilots who trained showed improved near acuity, contrast sensitivity and processing speed.
6 · Limitations, honestly
- Nearly all the presbyopia evidence comes from the method's developers, and several studies were funded by the company that commercialised it.
- No study used a sham-training control (people doing a similar but inert game). Test-retest learning, expectation and motivation (trying harder on the final chart) could explain part of the gain.
- Samples were modest, follow-up was short, and the main outcome is chart acuity under ideal conditions, not everyday reading comfort.
- Training does not stop the lens from aging. Many people, especially with readers stronger than about +2.00, will still need readers for small print.
- Home screens vary in brightness and gamma. Kestrel's contrast numbers track your progress on your screen; they aren't clinical measurements.
Our take: the mechanism is credible, the effect sizes are meaningful, and the cost and risk are near zero. Treat it as a promising experiment on yourself. Test before and after, under the same conditions.
References
- Polat U, Schor C, Tong J-L, Zomet A, Lev M, Yehezkel O, Sterkin A, Levi DM. Training the brain to overcome the effect of aging on the human eye. Sci Rep 2:278 (2012). doi:10.1038/srep00278
- Lev M, Ludwig K, Gilaie-Dotan S, et al., Polat U. Training improves visual processing speed and generalizes to untrained functions. Sci Rep 4:7251 (2014). doi:10.1038/srep07251
- Yehezkel O, Sterkin A, Lev M, Levi DM, Polat U. Gains following perceptual learning are closely linked to the initial visual acuity. Sci Rep 6:25188 (2016). doi:10.1038/srep25188
- Polat U, Sagi D. Lateral interactions between spatial channels: suppression and facilitation revealed by lateral masking experiments. Vision Res 33:993–999 (1993).
- Polat U. Making perceptual learning practical to improve visual functions. Vision Res 49:2566–2573 (2009).
- Sterkin A, Levy Y, Pokroy R, et al., Polat U. Vision improvement in pilots with presbyopia following perceptual learning. Vision Res 152:61–70 (2018).
- Allard R, Faubert J. The noisy-bit method for digital displays: converting a 256 luminance resolution into a continuous resolution. Behav Res Methods 40:735–743 (2008).
- Ben-Haim Sitbon M, Lev M, Polat U. Binocular fusion disorders impair basic visual processing. Sci Rep (2022).