The Autorefractor: What the Machine Measures Before You Answer a Question
A measurement taken without asking you anything
An autorefractor is the instrument that produces the little printout of numbers at the start of an eye examination. Its appeal is obvious: it takes a few seconds, it requires no answers, and it is repeatable. Its limitation follows from the same fact. Because it never asks you anything, it cannot know whether the number it computed is the number you will actually tolerate on your face, and those are not always the same.
Understanding what the machine does physically explains both the usefulness and the failure modes, and it makes the difference between the printout and your final prescription look like optics rather than mystery.
The optical principle: find the plane where the retinal image is sharp
Every autorefractor does one thing: it establishes the refractive state of the eye by measuring how a known pattern of light returns from the retina. A target — classically a ring or a set of mires — is projected into the eye along the instrument's axis. That target is imaged on the retina, and the light diffusely reflected from the retina travels back out through the eye's optics carrying the eye's defocus in the shape of its blur.
The instrument then changes the vergence of the returning beam, either by moving the target along the axis or by inserting lenses, until the detector sees a sharply focused image. The amount of adjustment needed is the defocus. Because the eye is not spherical, the measurement is repeated along several meridians; a modern instrument samples many points across the pupil at once using a lenslet array and reconstructs the wavefront, which is then reduced to three numbers: sphere, cylinder and axis.
This reduction is a fit, not a photograph. A wavefront with coma or trefoil still has to be expressed as a sphero-cylinder, which is why the machine's output can be perfectly reproducible and still not describe the aberration that bothers you at night.
What the printout actually reports
- Sphere, cylinder and axis, usually to the nearest 0.25 D and the nearest degree.
- Pupil size and pupillary distance, measured under the instrument's infrared illumination — a dim setting, so the pupil figure tends to be larger than in a bright room.
- Keratometry values on most instruments: corneal curvature radii and the corneal cylinder, which are useful for contact lens fitting and for separating corneal from lenticular astigmatism.
- Vertex-referenced power. By convention, the numbers refer to a plane about 12 mm in front of the cornea. This is the detail most often missed: a refraction of −6.00 D at a 12 mm vertex is not −6.00 D at the corneal surface, which is why a contact lens prescription for the same eye carries a different, smaller number.
Why the instrument tends to over-minus
The single largest error source is accommodation. The instrument sits close to the face, the target is real, and a young eye in particular will focus on it. When the eye accommodates during the measurement, the machine reads the accommodated state and reports more myopia than the eye has at rest. This is instrument myopia, a cousin of the night myopia that affects dim-light vision.
Instruments fight back by fogging: they deliberately add plus power so the target is slightly beyond the eye's far point, which encourages accommodation to relax. It helps, and it does not eliminate the problem, particularly in children, in early presbyopes whose accommodation is slow to release, and in hyperopes who accommodate habitually without noticing.
Other error sources are physical. A cataract or a corneal scar scatters the returning light and can shift the reading; a broken tear film does the same intermittently, which is why a reading taken immediately after a blink can differ from one taken a second later; a small or irregular pupil truncates the sampled wavefront; and soft contact lenses left in the eye change the anterior surface being measured.
What the machine cannot do
It cannot balance the two eyes against each other. It measures each eye separately and reports each separately, and the final prescription is usually adjusted so that the two eyes are balanced for comfort rather than each being optically optimal in isolation. It cannot determine whether you will accept a cylinder axis that is technically correct but unfamiliar. It cannot assess binocular vision, fixation disparity, accommodative facility or the health of the retina, optic nerve and anterior segment. And it cannot measure visual acuity, because acuity depends on neural processing, not only on focus.
Agreement between autorefraction and a careful subjective refraction is often within a quarter of a dioptre in straightforward eyes. The cases that fall outside that band — high cylinders, large anisometropia, children, post-surgical eyes, anyone with a strong accommodative response — are precisely the cases where the difference matters.
Five things patients say that are not quite right
"The machine can't be wrong, it's a computer." The machine is accurate at measuring what it measures. What it measures is a snapshot of a living, focusing eye in an unnatural viewing condition.
"The printout is my prescription." The printout is a starting point for refinement. Ordering glasses from it skips the subjective stage, the binocular balance and the tolerance check that turn a measurement into a prescription.
"The autorefractor checks my eye health." It measures refraction and, on most models, corneal curvature. It does not image the retina or measure intraocular pressure.
"Two machines gave different numbers, so one is broken." A quarter of a dioptre of difference between instruments is ordinary, and a difference in the cylinder axis of a few degrees is common in low cylinders.
"If the reading changed, my eyes got worse." A change of 0.25 D between visits is usually measurement noise, tear film state or accommodative tone, not progression.
When the printout is not enough
Ask for a full subjective refraction, rather than accepting the automated reading, if any of these apply: your two eyes differ by more than about a dioptre; you have a cylinder above about 1.50 D, where axis errors become visible; you are a child or a teenager, where accommodation is strong and myopia control decisions depend on accurate numbers; you have had cataract or refractive surgery; you are over 40 and your reading distance has changed; or you have symptoms the numbers do not explain — halos, starbursts, fluctuating vision through the day, or headaches that appear only with the new pair. Those are reasons for a proper examination, not for a different pair of single vision glasses ordered from a printout.