[visionlist] Teller Acuity Cards

Gislin Dagnelie gislin at lions.med.jhu.edu
Wed Feb 19 18:41:44 -04 2020

Thanks, John.

As my old teachers Van der Tweel and Spekreijse used to say:  "Detection is a 
nonlinear process."  So yes, even the most perfectly balanced high-resolution 
stimulus will become unbalanced due to neural rather than optical factors.

But as you say, it doesn't matter in the end: Vision is inherently non-linear 
and our visual acuity,whether for optotypes or for spatially periodic stimuli, 
depends on that.  But you're right that nonlinear processing of a luminance- 
balanced textured stimulus could result in the contour being visible by virtue 
of the brightness difference between pattern and background caused by the 
on-off asymmetry.

Tapering contrast along the contour edge would reduce the visibility of the 
luminance step, so in an ideal world that should be done to make the stimulus a 
true resolution test.  It can certainly be done through dithering in on-screen 
tests, and should be, since nonlinearity of the screen gamma function can 
further complicate matters.

Oh yes, and the occurrence of Moiré patterns for fine gratings is a good reason 
to use checkerboards or 2-D random dot textures instead, but that's an old 
hobby horse of mine.


On 19 Feb 2020 at 12:10, John Robson <jgr11 at cam.ac.uk> wrote:

Ruth et al,

Teller and other similar acuity tests.

Unless I have missed something, all those who have contributed so far to 
this thread seem to have assumed implicitly, or in some cases explicitly, 
that the patterned area will be indistinguishable from the uniform area 
when the average luminance of the two areas is the same. While there has, 
very reasonably, been some discussion about the conditions necessary to 
prevent the contour between the two areas being independently visible, the 
basic requirement of equal mean luminance has not been questioned.

If the spatial contrast in the retinal image were to be reduced to below 
visual threshold by a linear process (e.g. simple optical blurring) I am 
sure that this would be a correct assumption. But this is not normally the 
case even in adults. Even in adults the retinal image of a grating at the 
acuity limit will still have some, albeit much reduced, spatial contrast so 
that some cones will be more stimulated than the average and some less than 
the average. As was shown long ago by Donald Macleod (see Macleod, Williams 
and Makous 1992 Vis Res 32: 347-363 and other papers quoting this one) 
cones are non-linear so that the average signal from an array of cones 
which are non-uniformly illuminated will not be the same as the average 
signals from the same cones when illuminated uniformly. In general the 
brightness of the patterned and non-patterned areas will only be equal when 
the luminance of the uniform area is slightly different from that of the 
average luminance of the patterned area.

In fact it is not only the cones that are non-linear; there are intensive 
non-linearities at every stage in the visual system which will all no doubt 
contribute to distortion of the visual signal to an extent that depends 
upon how spatially non-uniform the neural signal is at that stage. While 
for adults it may well be that the non-linearity of the cones is the most 
important factor in affecting the apparent brightness of invisibly 
patterned areas with spatial frequencies close to the resolution limit of 
the eye's optics, this may well not be the case for visible patterns of 
lower spatial frequency whose apparent brightness is probably dependent on 
later non-linearities. It may well also not be true in infants in whom 
development of acuity lags behind development of their eyes' optics so that 
the retinal contrast of patterns with elements close to their acuity limit 
may have quite high retinal contrast.

I am not aware of any published report of the experimental determination of 
the apparent brightness of grating patterns with spatial frequencies above 
the acuity limit but I wonder if Ruth may not just have reported such 
observations. It is my understanding that Teller charts made by Precision 
Vision have the reflectance of the uniform area matched to that of the 
grating by visual inspection of the charts at a distance at which the 
grating pattern is invisible to an adult observer. I believe that they do 
this because they found that instrumentally matched luminances did not 
result in perceptual matches. I should be very interested to hear whether 
the charts on which Ruth made luminance measurements provided perceptual 
brightness matches. I have no idea whether any of the other similar tests 
use instrumental or subjective matching but for the sake of uniformity I 
think the matches should be made instrumentally. It is not at all certain 
that an adult match would be the same as that of an infant.

Does any of this matter? I am with Christopher on this. The test is what it 
is. Does it matter if the patterned patch is seen by virtue of its 
different apparent brightness, by having an illusional contour or because 
the pattern itself is visible. My answer would be no. We are discussing a 
clinical test here and what is important is whether the test is practical, 
reliable and useful. It is always important to remember that no two tests 
that are not physically identical can be expected to give exactly the same 
numerical answer and the existence of norms for each particular test is 
obviously an important consideration.

As to Ruth's observation that some subjects have anomalously high acuity 
with charts of this kind, it is not generally appreciated that the acuity 
limit for a grating may be determined by two different criteria: 1) can the 
bars of the grating be resolved well enough for the subject to determine 
their direction? or 2) can the subject distinguish a grating from a 
non-patterned area? It is particularly obvious when interference fringes 
are viewed that when the spatial frequency is raised above the level at 
which the direction of the bars can be clearly seen and reported the 
grating is still easily distinguishable from a uniform area. This is a 
result of the irregular moire pattern that is set up by interaction of the 
grating with the retinal cone mosaic (see p723 of Byram 1944 JOSA 34:718 
and Campbell & Green 1965). In subjects with particularly good optics this 
can be seen with regular imaging.

Hope this helps rather than confuses.


More information about the visionlist mailing list