[visionlist] observer rejection query

Fri Feb 12 17:47:59 GMT 2010

A few weeks ago,  I sent this query round the visionlist community.
Replies are summarised below.  Thanks to all who contributed.

Julie

-------------------------

A query to all psychophysicists, and especially to those working in binocular
vision.

We are working on a study using a large number of totally naive observers
(dozens of people rather than 3-4).  We're finding that many cannot achieve
threshold in a binocular vision task, sometimes around 25% of people.

That sounds like a lot, until you start sampling the psychophysics literature,
where binocular vision (and other) studies often report that 'observer X was
rejected because they could not do the task'.  With just a handful of
observers, losing 1 _could_ really reflect 25%.  Question is: HOW MANY people
typically can't do these tasks?  I'd like to work out a reasonable expectation
for the proportion of observers who deliver unusable data (whether reported in
publications or not).

I'd be very pleased if you could let me know your own experiences from your lab
studies (or others you may know of).

I will summarise any responses and post on visionlist.

regards

Julie Harris

Replies:
Suzanne McKee

In two weeks, I am submitting a paper describing 
exactly this problem.  I had subjects with 
exquisite stereoacuity for real metal bars; in 
the stereoscope, their thresholds were dreadful, 
if we could get them to do the task at all.  A 
draft of the paper is attached.

Marty Banks

We find that a lot of inexperienced observers can't initially see depth in
random-element stereograms. They see them as flat. Not sure what
percentage, but I wouldn't be surprised if it's 25%. With practice, the
great majority learn to see depth just fine.

Nearly all inexperienced observers can see depth in stereo photographs
where other depth cues are consistent.

Terri Lewis

We test large numbers of visually normal adults 
and children and, on clinical tests (WIRT and 
Randot) most have normal binocular vision. Of 
course, normal on a clinical test is 40x above 
thresholds tested psychophysically under ideal 
conditions. We also include in our study, only 
those with no history of eye problems. Any lazy 
eye or other childhood eye problems will result 
in reduced binocular vision.

Zhaoping Li

I observe similar things, about 25% of my 
subjects can not tell the which of the two 
transparent parallel depth planes (made of 
textures) are in front or back. But then, when I 
made a few objects in different depths and ask 
observers to tell me which ones are in front and 
back, I notice that fewer people are unable to do 
them.

David Ferster

It's been a long time since I've worked on 
binocular vision, but when I was doing my thesis 
work, I read and referred to a paper by Whitman 
Richards in which -- if I remember correctly -- 
he said that about 25% of the population had some 
deficit in stereovision. Some could not tell when 
things were farther from the horopter, others 
when things that were nearer. Others were 
completely stereoblind.

RICHARDS, W. (1971). Anomolous stereoscopic depth 
perception. J. opt. Soc. Am. 61, 410-414.

Branka Stirn Kranjc

Thank you for your interesting information, I do 
not think that you have got a false high result, 
though it depends very much on the used methods. 
I know the data from our outpatient clinic used 
for labour medicine (Rodenstock R7 apparatus) - 
being aware that it is not for a routine 
ophthalmological examination, but according to 
that test over 75% e.g.bank employees and also 
others failed in stereo testing. So we really 
function with low binocular vision and your 
observation is very true I believe.

Bart Farrell

I'd say 25% might be about right as some sort of 
average.  In my experience much of the 
variability around this average is 
stimulus-dependent.  In the scenario I'm most 
familiar with, people who do well with 
sharp-edged stimuli in screening tests fail when 
they have to see depth with blurry stimuli.  They 
seem to add retinal images without fusing them 
unless there are edges, with the result that 
disparity modulates the perceived contrast of 
sinewave gratings, for example.  Most of those 
who fail this way are 'stereo normal' according 
to standard tests and RDSs.  I've had mixed (but 
mostly negative) success in bring them around 
with practice.

Coutant & Westheimer (Ophthalmic Physiol Opt 
1993, 13:3-7) give much more optimistic 
estimates--maybe the trick is to use biology 
students, as they did.

Jenny Read

I haven't had experience with a /huge/ number of 
naive observers, but in the last couple of years 
we must have tested 40 or so naive observers, 
ranging from 4 years to 60. So far everyone 
without known visual problems has been 
immediately able to discriminate depth sign in a 
random-dot stereogram depicting a disparate disk 
on a zero-disparity background; we've been able 
to record a threshold in a few minutes using 
front/back discrimination with a staircase 
procedure. The only exceptions (two come to mind) 
are people with a known history of binocular 
vision problems, e.g. strabismus/amblyopia, and 
they've told us about their stereo problems 
before viewing the stimulus.

If I can get even more anecdotal, I almost always 
show a random-dot disk-on-background in my talks, 
esp. to audiences unfamiliar with stereo vision, 
and invite people to swap the glasses around and 
see how the depth inverts. I have always had the 
impression that this is working for almost 
everyone; I'd be very surprised/disappointed to 
learn that a quarter of my audience are not 
getting the effect!

However, all this is from using a passive-stereo 
projection system and/or anaglyph, so in both 
cases using 3D glasses. My impression is that 
mirror stereoscopes are harder because you have 
to get the correct eye alignment. Could this be a 
factor?

In other tasks, our observers (naive and 
experienced) have definitely had more problems. 
Tasks where I've failed to get a threshold on one 
or more observers include ones with an 
interocular delay, with pure absolute disparity 
(no reference surface to provide relative 
disparity), with vertical disparity and where you 
are asked to make stereo judgments in the 
periphery, difficulties compounded by short 
presentation durations. However, I regard all of 
these as "hard" tasks and an inability to perform 
well on them is not the same as not having stereo 
vision -- in every case where I've had to reject 
an observer (sometimes me!) on one of these 
"hard" tasks, the observer has had good 
stereoacuity on a nice simple front/back RDS task.

Arash Sahraie

We have run a number of studies on detection of a 
change in disparity in dynamic stimuli. We first 
screened for normal stereopsis using a standard 
static optometry test. We found that 
approximately 25% of around 60 naïve observers 
who had normal stereopsis had to be excluded as 
they were unable to detect a change in disparity.

Wolf Harmening

we are doing a simple stereo threshold task in 
one of our undergraduate  lab courses to 
demonstrate binocular vision and hyperacute 
resolution.  The task is to discriminate an 
object vs hole configuration (positive vs. 
negative disparities of a central square against 
zero disparity of background) in an anaglyphic 
RDS stimulus. The lab course runs once a year 
with a total of around 120 participants. I am 
overseeing this course for 5 years now.

Although I have not quantified this, your 
observation - 25% of subjects cannot achieve 
useful data - sounds perfectly congruent with my 
experience here. In a daily group of 12 students 
we are testing, usually 2-4 cannot do the task 
(i.e. data do not produce a psychometric 
function).

Raymond van Ee

~30% of normal observers is stereoanomalous.
~2% is stereoblind

Stereoanomalous means that they are unable to 
process either crossed or uncrossed disparities. 
This group needs to make well-adjusted eye 
movements to move the disparities from a visual 
scene within receptive fields that they are able 
to use.  This requires some learning or coaching.

In fact there is evidence that performance in a 
stereo-experiment correlates with the extent of 
stereoanomaly. It is easy to run a stereoanomaly 
test. There is an example on my web site: 
http://www.phys.uu.nl/~vanee/RvE_SWare.html

literature:
W. Richards (1971). Anomalous stereoscopic depth 
perception. Journal of the Optical Society of 
America, 61, 410-419
R. van Ee (2003). Correlation between 
stereoanomaly and perceived depth when disparity 
and motion interact in binocular matching. 
Perception, 32, 67-84.

Jean-Jacque Sacre

As studying and developing future 3D cinema and 
3DTV we are investigating 3D perception and its 
variability.
Recently we conducted tests related to 3D vision 
with about 25 test persons. These tests were 
designed to better understand the visual 
discomfort that could exist when the perspective 
is not in agreement with the depth given by a 
stereo pair. The test also included a part to 
have information on the accommodation convergence 
conflict.

The observers were in majority image processing 
R&D persons, and we noted a proportion close to 
25% providing unreliable results. In total 4 to 9 
observers gave results difficult to exploit at 
various degrees. Among them, 3 observers reported 
having a poor 3D vision to their 
knowledge/opinion.

In our investigations we are interested to know 
what are the conditions to improve this score, 
and if specific ('better') stereoscopic or 
multiview presentation of images can help people 
to reporting difficulties.

Laurie Wilcox

We have been testing naive adults for a 
developmental study (as a comparison group for 
5yr olds). We've found enormous variability in 
performance, and a large number of people who 
barely pass the Randot stereotest. Of those who 
do pass the conventional tests, there are a 
number who cannot seem to do the psychophysical 
tasks. Of course, this could be due to many 
factors unrelated to stereopsis, and with more 
experience they would likely improve.  I would 
say that your estimate of approximately 25% of 
observers not being able to do stereo tasks 
without practice is a reasonable one. The 
question remains if this is because they need to 
'learn' how to use stereo in isolation (something 
that doesn't happen often in the natural world), 
or if they are truely stereo deficient.

Walter Makous

No doubt the proportion of stereoblind observers 
depends on a lot of variables.  However, Bela 
Julesz showed a lot of random dot stereograms to 
a lot of people, and I recall his saying about 
10% reported not being able to see them.  (I 
think he also stated that in his book, 
Foundations of Cycloplean Perception, but I since 
moving out of my office I can no longer find my 
copy.)

I've shown a fair number of stereograms of 
varying type, including the random dot variety, 
mostly to undergraduate classes.  I always asked 
how many viewers could see the stereo, and 
although I did not keep any data on it, but my 
impression is that the proportion of stereoblind, 
by this test, was much closer to 10% than 25%.

Cliff Schor

Richards and Regan used to study stereo visual 
fields and found that regions sensitive to motion 
in depth were very sparse and idiosyncratic. 
Usually there was sensitivity near the fovea but 
if your stimuli are peripheral, that could 
account for the failure of stereo in some people. 
Also 3% of the population has some serious form 
of binocular anomaly and at least 5% have ocular 
alignment problems stemming form high phorias or 
refractive errors that produce conflicts between 
accommodation and covergence.  As Marty said, a 
lot of people can be trained to improve, See 
Westheimer and Fendick, and don't do well in 
stimulus conflict conditions.  I once did a 
screening with stereograms and was surprised how 
many people did not see the obvious depth. 
Frankly, I'm more surprised how many people like 
the magic eye stereograms when they are so 
difficult to fuse.  It shows a lot of success has 
to do with motivation but begs the question of 
what happens in normal circumstances when the 
motivation is not there.  This could be a serious 
problem for the new 3_D movie and TV industry (a 
big headache).

Jeannine Pinto

I don't do research in binocular vision but I do 
classroom exercises that require students to fuse 
random dot stereograms.  In a class of 10-15, I 
typically have one or two who can't do it, even 
after repeated attempts to train them.  A couple 
of years ago, less than half of the group was 
able to do it.  All in all, I've probably had 
nearly 20-25% of the students unable to fuse them 
stereograms.

Richard Hetley

Hello.  This sort of question is exactly what I 
would like to see addressed in more detail. 
Outside of a laboratory setup, when I am 
demonstrating binocular phenomena to naive 
observers in a small or large crowd by means of 
filters, glasses, and suchlike, an inevitable 
response is "Nope, I can't see that one" followed 
shortly by "Well, if people can't see your 
effect, what does that tell you?"  Well, what 
does it tell us?  The same for when naive 
observers do come into the lab, do experience a 
controlled setup, and still have some effects 
work while others do not.  Can we, in fact, come 
up with patterns in the proportion of individuals 
who cannot see one or another binocular 
phenomenon, yet who still take advantage of 
binocular information in the global sense?

I hope there is more rigorous study of this out 
there somewhere, but for now . . . to add to your 
pool of "3-4 person studies," I did two batches 
of those.  The first batch of experiments 
required binocular fusion of images and 
perception of the resulting depth, luminance, and 
luminance contrast.  Three people were successful 
and one person was dropped, the latter due to 
failure to see any effect in the fused images. 
The second batch involved (binocular fusion of 
images and) perception of depth, binocular 
luster, and binocular rivalry.  Three people were 
successful and one person was dropped, which 
should sound familiar, but note that two of the 
three successful participants were held over from 
the first batch.  In total: four unique 
successful participants, two dropped.

David Brainard

We screen observers for their ability to use 
binocular disparity before running them in
experiments on interaction of object color, 
material, and shape.  We use a commercial vision 
testing device that assesses discrimination for 
four levels of disparity.

Our device is a Keystone VS-II 
(http://www.keystoneview.com/).  It doesn't look 
like they make this exact model anymore.  It's 
stereo acuity test is calibrated on the 
Shepherd-Fry scale, with passing corresponding to 
a disparity that they list as 75% on this scale. 
16 of 17 observers passed at the criterion given 
on the device, corresponding to 75% on the 
Shepherd-Fry scale, once we adjusted it correctly.

Arthur Lugtigheid

We have a disparity screening test that we use 
for naive participants. It's a simple near/far 
disparity discrimination (range +/- 9 arcmin) 
task using random dot stimuli presented for a 
limited duration. Prompted by your message, I've 
reviewed the data collected in the lab from naive 
participants over the past year (84 in total).

We find 23/84 participants have psychometric 
functions where the standard deviation of the 
fitted Cumulative Gaussian is > 2.5 arcmin - 
which for our purposes is unacceptable. Of these, 
a number of participants (11) appear to respond 
on the basis of some disparity being present, 
regardless of its sign (a 'V' or inverted V 
pattern in the psychometric functions).

For a subset of participants (13) we have data 
from two attempts. Seven of those (~55%) show 
poor performance in both. Six (~45%) initially 
show poor performance but then show good 
performance in the second attempt.

Gary Rubin

We did a study of a random sample of 2520 older 
adults (65-85) and tested stereoacuity with 
Randot Circles. We found that a sizeable 
proportion of the population could not see the 
largest disparity ("stereoblind"). Of course this 
included people with poor vision in one or both 
eyes, which we discuss. I have attached a copy of 
the paper from IOVS. A relevant excerpt is given 
below.

"The prevalence of stereoblindness increased 
significantly with age (*2 = 64.9, P < 0.0001) 
from 10% in the 65 to 69 age group to 26.3% in 
the 80 to 85 age group. The age-adjusted 
prevalence of stereo- blindness was higher for 
men than for women (16.1% versus 13.0%, x* = 4.6, 
P < 0.05) and higher in black participants 
compared to white participants (19.0% versus 
12.7%, x2 = 15.8, P < 0.0001)."

-- 

=========================================================

Julie M. Harris
Prof. of Psychology
St. Andrews Vision Lab
School of Psychology
University of St. Andrews
St. Mary's Quad
South St.
St. Andrews
KY16 9JP

tel:  44-1334-462-061
fax: 44-1334-463-042
email: Julie.Harris at st-andrews.ac.uk
http://www.st-andrews.ac.uk/%7Ejh81/vislab.html

The University of St Andrews is a charity registered in Scotland : No SC013532

=========================================================
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://visionscience.com/pipermail/visionlist/attachments/20100212/50547571/attachment-0001.htm>