I use a streak retinoscope at the beginning of every refraction.  By
blurring the patient with plus and moving the light across the retina
in various meridians, I can estimate the refractive error to within
about 0.50 diopters.  It takes less than five minutes for both eyes.
The autorefractor uses the same principal, except that the light is
non-visible infrared.

I once had a patient who was a myope of -2.50 diopters, and who wore
contact lenses.  Two years later she came back with a prescription of
-7.50 in eyeglasses.  As a nurse, she had developed some type of
allergy/infection and had her eyes examined at the medical school.  She
was given a prescription from the autorefractor "to get her by" until
she could wear her contact lenses again.  Despite the fact that she had
worn the eyeglasses for nearly one year, she still had a manifest
refraction of -2.50 in my office.

The take home message is that the autorefractor was capable of a 5
diopter magnitude error.  This is why it is important to check the
results.  On the other hand, the standard deviation is probably closer
to +/- 0.50 diopters.

The other message for "Otis fans," is that over-minusing a patient by 5
diopters did not increase the patient's myopia.

DrG

Not really; you could use an autorefractor after cycloplegia and perhaps get
a better result.

The image faded in and out because the instrument "teases" your
accommodation to relax by focusing the image further and further away.

Autorefractors have improved greatly in the last 20 years but they're still
way off, sometimes. A beam of infrared is shone into the eye and focused on
the retina. The instrument "tunes" the focused image on the retina and
reaches a nominal "refraction."

Autorefractors (and retinoscopy) are considered "objective" because you
don't have to participate.  The "1 or 2" refraction is called "subjective"
because the subject chooses the final value. "Manifest" is an old term,
coined before there were autorefractors. In current usage it's synonymous
with the "subjective" refraction, with or without cycloplegia.

Autorefractor results are now more _consistent_ from trial to trial. But
even when the results are identical, over and over, it will vary somewhat
from the "subjective" refraction. Retinoscopy is an old technique that works
similarly - the results can be very consistent but still not match the final
"1 or 2" refraction. These days most myopia research is done with
autorefractors and cycloplegia. The reasoning is that even if the result
doesn't match the subjective refraction, consistency is more important if
you're trying to measure change over time.

All these instruments have their own tendency to error. So does the
subjective refraction. What's most useful is comparing the two (or three)
different types of measurement to know what a single individual is doing.
Autorefractors (and retinoscopy) tend to show astigmatism that isn't
confirmed on the subjective refraction. Each instrument has its own tendency
to cause "instrument myopia" where a person simply accommodates more when
you stick his head in an instrument or hold something up to his face. That's
where you get the 5-diopter errors. However it's much more common to see
large errors in farsighted people than in nearsighted folks. Farsighted
people generally have a habit of accommodating constantly. Getting the
_accurate_ result often requires multiple doses of cycloplegia. However,
prescribing the _accurate_ result often makes them come back mad - they
throw their glasses down on the counter saying "these don't work!"

That's why I usually use the subjective refraction _without_ cycloplegia for
the eyeglass prescription. It's done in the "habitual state" and the lenses
are essentially chosen by the patient. If there's a problem with the
glasses, say they're made too minus, it's because the patient told me that
any less minus was blurry.

Otis is fixed on the idea that minus causes harm, so he believes every
myope's glasses should be made blurry and making them sharp should be called
"overcorrection."  In the real world, glasses are made with the minimum
minus necessary for best acuity. Anything less is "undercorrected" because
roadsigns are blurry. Anything more is "overcorrected" because it doesn't
add acuity.

It used to be popular to believe that overcorrecting would hasten myopia. In
fact it doesn't, except by stimulating a little more accommodation, the same
way farsighted people do constantly.

Likewise you might think that undercorrecting would slow the progress of
myopia, but it actually seems to do the opposite - witness the Indonesian
study that had to be terminated because the undercorrected group got worse
_faster_.

-MT