A Behavioral Method For Efficient Screening of Visual Acuity in Young Infants
A Behavioral Method For Efficient Screening of Visual Acuity in Young Infants
A Behavioral Method For Efficient Screening of Visual Acuity in Young Infants
A technique for rapid behavorial screening of grating acuity in infants 1 to 4 months of age is
described. The approach, called forced-choice preferential looking (FPL), depends upon the
fact that normal infants will stare fixedly at acuity gratings with stripe widths above a rather
abrupt cutoff width. The present task is to find the minimum stripe width, termed the diagnos-
tic stripe width, to which infants with normal visual acuity will readily respond. The expecta-
tion is that infants with below-normal acuity will not be able to respond to diagnostic stripes so
defined. To assess the feasbility of such a test procedure and define preliminary diagnostic
stripe widths for infants 4, 8, 12, and 16 weeks of age, 76 presumptively normal infants were
tested in the laboratory with a 40-trial FPL procedure. Sixty-nine infants (91%) completed the
test procedure. For each age group, a preliminary estimate of the diagnostic stripe width was
found. The group performance was uniformly high for the stripe widths designated as diagnos-
tic and fell off sharply for finer stripes, confirming the feasibility of the approach.
1142 0146-0404/78/121142+09$00.90/0 © 1978 Assoc. for Res. in Vis. and Ophthal., Inc.
Fig. 2. Observer's view of the infant through the peephole. A, Acuity grating to the observer's
left. B, Acuity grating to the observer's right.
80 MIN.OFARC
(20/1600) (SNELLEN)
I OCTAVE
STRIPE WIDTH
Fig. 3. Psychometric function for an 8-\veek-old infant tested with the FPL procedure. The
observer's percent correct is plotted as a function of the five stripe widths on which the infant
was tested. Conversions to Snellen equivalents (using the convention that 1' per stripe equals
20/20 Snellen) are given on the abscissa. The infant performed near 100% for 40' and 20'
gratings but near chance for 10' and 5' targets, with an estimated "threshold" (75%) at about
16' or 20/320 acuity.
several different stripe widths is required to obtain with a highly efficient search strategy, many stripe
a well-defined psychometric function. To obtain a widths would have to be sampled to find and
psychometric function with 20 trials per point at define the function if it were in an unpredictable
five stripe widths can take from 2 to 4 hours of location.
testing, depending upon the infant. Second, if the A procedure which should on the average be
stripe widths needed to span the function cannot less time consuming would be to test an infant at a
be guessed a priori—and, almost by definition, single point on the stripe-width continuum and
they cannot be for clinically interesting infants— then use the infant's score on that stripe width as a
then even more stripe widths must be used. Even diagnostic indicator of whether the infant's acuity
is normal or not. Ideally, such a diagnostic stripe Stripe widths of the grating stimulus were 0.95,
width would be selected to be large enough so that 0.48, 0.32, 0.24, and 0.12 cm, which correspond
virtually all infants with normal visual acuity could to approximately 80', 40', 27', 20', and 10' of vi-
easily detect the stripes but small enough that vir- sual angle, respectively, for an infant held at the
tually all infants with abnormally low acuity would test distance of 36 ± 3 cm from the centers of the
fail to detect them. For clinical evaluations of in- two stimulus positions. Conversions to Snellen
dividual infants, in addition to using a minimal equivalents are given in Fig. 3.
number of trials, it is also necessary to achieve Three adults are required to test an infant: the
statistically reliable results for each infant. observer, who observes the infant and judges the
With these strategies in mind, we have defined position of the striped card on each trial; the
the diagnostic stripe width to be the narrowest holder, who holds the infant but cannot see the
stripe width on which 95% of normal infants of a stimulus display (Fig. 1); and the experimenter,
given age will get 15 or more out of 20 trials cor- who sets up the trials, records the observer's
rect; i.e., 75% or more correct on 20 trials.* This judgments, and provides trial-by-trial feedback to
value was chosen because with p = 0.5 and the observer.
n = 20, 75% correct is statistically significantly Subjects. Subjects were solicited by letters to
above chance at the 0.05 level (exact binomial parents listed in the birth announcements section
test). of the newspaper and were paid $5.00 for the test
The present task, then, was to find and validate session. Fourteen 4-week-old, twenty-one 8-
diagnostic stripe widths for normal infants of vari- week-old, nineteen 12-week-old, and twenty-two
ous ages. 16-week-old infants were tested. All infants were
Apparatus. The apparatus was identical to that born within 10 days of their due date, as reported
described by Teller et al. 5 and is shown in Fig. 1. by the parents, and each was tested at a single age,
Basically, the apparatus consisted of a gray screen within 3 days of his or her fourth-, eighth-,
(Crescent Cardboard No. 651) in which two 9 cm twelfth-, or sixteenth-week birthday.
stimulus holes and a 4 mm peephole were cut. On Selection of stimuli. The goal of the study was to
each trial, an acuity grating (Biometrics Division, estimate, to one octave* or better, the diagnostic
Narco-Biosystems) was placed behind one stimu- stripe width, i.e., the smallest stripe width on
lus hole and a piece of the gray cardboard was which 95% of infants with normal visual acuity
placed behind the other. The acuity gratings and would perform significantly above chance. There-
gray cards were embedded in a wheel mounted fore, for each age it was necessary (1) to find a
behind the cardboard screen so that the position of stripe width on which virtually all infants would
the stripes (left vs. right) and the size of the stripes perform significantly above chance in 20 trials and
(control stripes vs. test stripes) could be changed (2) to show that a substantial fraction of normal
rapidly during the test session. infants would fail to perform significantly above
The acuity gratings had a nominal 1:1 duty chance on a stripe width finer by 1 octave or less.
cycle and a nominal contrast of 82%. The space- Intuition and previous experience in testing in-
average luminance of the individual striped cards fants5' l5 (Fig. 3) led us to use 40' and 20', 27' and
ranged from 1.25 to 1.30 log cd/m 2 (measured with 20', 20' and 10', and 20' and 10' test stripes for 4-,
an International Light IL500 Photometer) and ap- 8-, 12-, and 16-week infants, respectively. Each
proximately equaled the luminance of the gray infant was tested with only one of the two test
cardboard stimulus and screen (1.23 log cd/m 2 ). stripe widths selected for his or her age group.
In addition to the test stripes, every infant was
*It must be emphasized that the diagnostic stripe width tested with a set of large, readily resolvable con-
is not itself an absolute measure of acuity, but falls about trol stripes, of 80' subtense. The control stripes
1.5 octaves below the FPL acuity at each age. (See Dob- were intended to serve two purposes. First, they
son and Teller.17) The 1- to 1.5-octave difference be- helped keep the infant interested in the proce-
tween the estimated diagnostic stripe width and the av- dure. Second, they indicated whether poor per-
erage acuity found for infants of a particular age is not
formance by an infant on the test stripes (if it oc-
surprising, given that the average psychometric function
for an individual infant typically spans about a 2-octave curred) was due to a behavioral problem, e.g.,
range5' l4- IS (See also Fig. 3.) A stripe width 1.5 octaves fussiness or drowsiness, or to limited acuity. High
wider than that required for theshold (75% correct)
would be near the top of the psychometric function. *An octave is a halving or doubling of stripe width or
Thus, an infant with acuity near normal for its age would spatial frequency, and a halving or doubling of the de-
be expected to do very well on the diagnostic stripes. nominator in Snellen notation (Fig. 3).
Name Holder
Birthdate Observer
Date Experimenter
Trial 1 2 3 4 5 6 7 8 9 10
Grating 1600 1600 1600 1600 1600 1600 1600 1600 1600 1600
Position R L R R L R L R L L
Score
Trial 1 2 3 4 5 6 7 8 9 10
Grating
D D 1600 D 1600 D D 1600 D D
Position R L R R L L L R L R
Score
T if 5/5, quit
Trial 11 12 13 14 15 16 17 18 19 20
Position R L R L R L L L R L
Score
f if 9/10, quit
Trial 21 22 23 24 25 26 27 28 29 30
Position R R R L R L R L L R
Score
Fig. 4. Scoresheet containing one of the orders of trials used for testing an infant with the
diagnostic stripe width. On the first trials, large control stripes are presented to allow the
infant to become familiar with the procedure and the observer to become familiar with the
infant's looking pattern. The final 30 trials are a mixture of 10 control stripes and 20 diagnostic
stripes. The size of the diagnostic stripe width used depends on the age of the infant. In the
present study, all 40 trials were used. Arrows indicate a set of possible cutoff points of termina-
tion of testing for increased efficiency. (See discussion of rescoring.)
performance (>75% correct) on the part of the performance on both test and control stripes
observer on both test and control stripes was taken (should it occur) would be taken to indicate either
to indicate both that the infant was testable and an untestable infant or one with severely limited
that the test stripes could be resolved. Poor per- acuity.
formance on the test stripes but high performance Procedure. Each infant received 40 trials using
on the control stripes was taken to indicate a test- the FPL procedure. The time required to com-
able infant unable to resolve the test stripes. Poor plete the 40 trials was recorded. A sample score
100 T
STRIPE WIDTH
Fig, 5. Percent correct on test and control stripes, for infants tested in preliminary estimation
of diagnostic stripe widths. Scores of each infant on a test stripe width and on the 80' control
stripes are shown. Preliminary diagnostic stripe widths are taken to be 40', 27', 20', and 20' for
4-, 8-, 12-, and 16-week infants, respectively. The three symbols representing scores on the
control stripes indicate scores of infants who passed (scored S: 75%) on either of the two test
stripes (•) and scores of infants who failed (scored < 75% correct) on the smaller (o) or larger
(x) of the test stripes at a particular age.
sheet is shown in Fig. 4. For the first 10 trials, the failed. In the latter case, no more infants were
control stripes were used. These served to ac- tested at that stripe width, since it was unlikely to
quaint the infant with the procedure and allowed be the diagnostic stripe width.
the observer to become familar with the infant's
looking pattern. Following these trials, the infant Results
received 30 trials in which 10 of the control stripes
were intermixed quasirandomly with 20 trials of The scores obtained by each infant on test
the test stripes being used for that infant. On each and control stripes are shown in Fig. 5. The
trial, the observer's task was to say whether the abruptness of the change in the infants' be-
stripes appeared on the left or on the right side of havior with changes in stripe width suggests
the screen. Trial length was variable, depending that the concept of a diagnostic stripe width is
upon the length of time the observer felt was re- a meaningful one. For example, for the
quired to make a judgment. Typically, trial length 4-week-old infants (Fig. 5), three out of three
was between 10 and 30 sec, although occasional infants tested with 20' stripes failed (scored
trials required less than 5 or more than 60 sec.
less than 75% correct), while 10 out of 10
An infant's results were designated pass if he or
infants tested with 40' stripes passed (scored
she scored 75% or better on the 20 test trials, and
fail if the score fell below 75%. greater than 75% correct). These data, al-
Testing on each potential diagnostic stripe though limited by a small sample size, sug-
width was continued either until 10 infants had gest that the diagnostic stripe width for
been tested with that stripe width (of whom at 4-week-old infants falls between 20' and 40'.
least eight had passed) or until three infants had Similarly, the data for the other age groups
suggest diagnostic stripes between 20' and fants, evenly distributed across ages. Two
27' for 8-week infants and stripes between 16-week infants did not complete testing
10' and 20' for 12- and 16-week infants. due to fussiness. For the 69 infants who
Therefore, for our preliminary clinical trials, completed all 40 test trials, the average time
we used 40' stripes for infants less than 7 required was 24 mins (SD = 10 min,
weeks of age, 27' stripes for 8- through 11- range = 10 to 45 min).
week infants, and 20' stripes for infants 12 Rescoring. It was often obvious to the ob-
through 16 weeks of age.22 server that infants who were going to pass
The results on the control stripes for the could see the diagnostic stripes long before
4-, 8-, and 12-week infants also support the the end of the 40-trial procedure. Thus, it
efficacy of the test procedure. For these age seemed sensible to stop testing each infant as
groups, virtually all infants scored 85% or soon as he or she obtained a score statistically
above on the control stripes (mean = 95%). significantly above chance. Binomial prob-
This performance suggests that any failure to abilities indicate that the probability that an
perform above chance on the test stripes was infant who cannot see the diagnostic stripes
probably due to real acuity limitations rather could obtain a score of 5/5, 7/8, 9/11, 11/14,
than insensitivity of the procedure. One 12- 12/16, or 14/19 is less than 0.04, with a cu-
week infant performed below 75% on the mulative probability of 0.088.22 If it is desired
control stripes; however, the fact that this in- to keep the cumulative probability at 0.05,
fant scored 95% on the 20' stripes suggests the cutoff criteria 5/5, 9/10, 13/15, and 15/20
that the test was an adequate assessment of may be used, for a cumulative probability of
that infant's visual acuity. 0.050.
Sixteen-week-old infants are, on the aver- Rescoring the data using either set of cutoff
age, a little harder to test. Some of these in- criteria showed that of the 46 infants who
fants became fussy toward the end of testing scored significantly above chance on the 40-
and objected to being held in front of the trial procedure, all would have passed, and
screen. The difficulties encountered in test- the average number of trials required to
ing this age group are reflected in the rela- complete testing in these infants would have
tively low scores of the 16-week infants on been 20 instead of 40, with a concomitant
the control stripes (mean = 89%). Neverthe- reduction of testing time. Of the 13 infants
less, two of the three 16-week infants who who failed to score significantly above chance
failed with the 10' stripes performed well on on the 40-trial procedure, only one (the 16-
the 80' stripes, suggesting that the test ade- week-old infant who failed the 20 min stripes)
quately assessed their detection of the test would have obtained a passing score if either
stripes. The third performed poorly on both, set of cutoff criteria had been used. Since this
suggesting lack of testability. The one 16- infant's poor score on the 40-trial procedure
week infant who failed on the 20' stripes was primarily due to fussiness during the
scored only 80% on the 80' stripes. By strict latter part of the test, the results using the
usage of our scoring rules, he would be diag- cutoff criteria are probably a more accurate
nosed as having failed on the basis of poor indication of the infant's abilities than the re-
acuity. However, his low performance coin- sults after 40 trials. Furthermore, five of the
cided with fussiness at the end of testing. seven infants who were eliminated from the
Diagnosis of limited acuity vs. untestability 40-trial procedure due to sleepiness or fussi-
would in fact have been unclear for this in- ness during testing would have been able to
fant, and retesting on a later date would be complete the version of the procedure con-
indicated. (See also the results of rescoring, taining the cutoff criteria. Thus, use of cutoff
below.) criteria to shorten the 40-trial procedure does
Seven of seventy-five (9%) of the infants not appear to change the conclusions drawn
failed to complete the test session. Sleepiness about an infant's vision in the laboratory; it
prevented completion of testing of five in- merely shortens test time to one half or less
for infants with good visual acuity and allows cal Center, Boston,22 in order to try the
testing of infants who would not tolerate a technique in a clinical as well as a laboratory
longer 40-trial procedure. setting. We wish to see whether infants at
risk for poor visual acuity fail to perform
Discussion significantly above chance on the diagnostic
The results of the present study suggest stripes, and whether infants who perform
that it is possible to find a reliable minimum poorly without prior diagnosis of risk will be
(diagnostic) stripe width on which most in- found to have significant visual problems.
fants of a particular age with presumptively Finally, it should be reemphasized that the
normal visual acuity can quickly perform present modification of the FPL technique is
significantly above chance during testing specialized as a screening test and not as a
with the FPL procedure. For each age measure of an infant's acuity per se. That is,
group, infants tested with the stripe width each infant is tested against a single age
later designated as diagnostic obtained uni- norm—the diagnostic stripes appropriate for
formly high scores, while infants tested with his or her age—and simply passes or fails; no
stripes 0.5 or 1 octave narrower showed a estimate of acuity is made. Acuity estimates
much greater variability of scores. can be made with FPL by using several stripe
On the basis of previous laboratory acuity widths in random order on the same infant
data on infants, it was possible to predict and generating a psychometric function simi-
fairly accurately which stripes would be the lar to that in Fig. 3, as in our usual laboratory
diagnostic stripes. This success suggests that procedure, 5 ' 14> 15> 26 or by the use of staircase
our quick assessment taps a visual function or staircaselike procedures.12" l3> 27 Attempts
very similar to that measured in more exten- are in progress in our laboratory and in the
sive FPL acuity testing. The validity of our laboratories of others to develop a preferen-
results is further supported by the parallel tial looking technique that is efficient and re-
fashion in which the size of the diagnostic liable enough to yield useful acuity estimates
stripe widths and previously reported esti- for individual infants in clinical settings.
mates of PL acuity16* 17 change with increas-
We thank Michelle Backholm and Michael Sekel for
ing age of the infant. Further, the test is re- assistance in running subjects, and Marjorie Zachow for
latively quickly accomplished, and more than secretarial assistance.
90% of the infants tested completed testing.
REFERENCES
Thus, the technique exhibits some of the
properties of efficiency and validity necessary 1. Gorman, J. J., Cogan, D. C , and Gellis, S. S.: An
apparatus for grading the visual acuity on the basis of
for eventual usefulness in the clinical setting. optokinetic nystagmus, Pediatrics 19:1088, 1957.
Further development of the procedure will 2. Dayton, G. O., Jones, M. H., Aiu, P., Rawson, R.
include several aspects. First, the size and A., Steele, B., and Rose, M.: Developmental study
diversity of the sample on which the diag- of coordinated eye movements in the human infant.
I. Visual acuity in the newborn human: a study
nostic stripes are selected will be increased, based on induced optokinetic nystagmus recorded
and finer estimates of the diagnostic stripe by electro-oculography, Arch. Ophthalmol. 71:865,
widths will be made. Second, increasingly 1964.
shorter and more efficient forms of the pro- 3. Catford, G. V., and Oliver, A.: Development of vi-
cedure are being developed, with the goal of sual acuity, Arch. Dis. Child. 48:47, 1973.
4. Fantz, R. L., Ordy, J. M., and Udelf, M. S.: Mat-
reducing the time required for screening
uration of pattern vision in infants during the first six
acuity to a maximum of 10 min. Third, mon- months, J. Comp. Physiol. Psychol. 55:907, 1962.
ocular testing of infants would presumably be 5. Teller, D. Y., Morse, R., Borton, R., and Regal, D.:
of considerable value in early detection of Visual acuity for vertical and diagnonal gratings in
monocular problems. We are presently at- human infants, Vision Res. 14:1433, 1974.
tempting to test infants monocularly. Fourth, 6. Harter, M. R., and Suitt, C. D.: Visually-evoked
cortical responses and pattern vision in the infant: a
infants are being tested in the Department of longitudinal study, Psychonomic Sci. 18:235, 1970.
Ophthalmology at Children's Hospital Medi- 7. Marg, E., Freeman, D. N., Peltzman, P., and
Goldstein, P. J.: Visual acuity development in and visual correlation of an infant bom with unilat-
human infants: evoked potential measurements, IN- eral eye lens opacity, Doc. Ophthalmol. 41:371,
VEST. OPHTHALMOL. 15:150, 1976. 1976.
8. Sokol, S.: Measurement of infant visual acuity from 19. Miranda, S.: Visual abilities and pattern preferences
pattern reversal evoked potentials, Vision Res. of premature infants and full-term neonates, J. Exp.
18:33, 1978. Child Psychol. 10:189, 1970.
9. Sokol, S.: Patterned elicited ERGs and VEPs in 20. Fantz, R. L., Fagan, J. F., and Miranda, S. B.:
amblyopia and infant vision. In Armington, J., Early visual selectivity as a function of pattern vari-
Krauskopf, J., and Wooten, B., editors: Visual Psy- ables, previous exposure, age from birth and con-
chophysics: Its Physiological Basis, New York, ception, and expected cognitive deficit. In Cohen,
Academic Press, Inc. (In press.) L. B., and Salapatek, P., editors: Infant Perception:
10. Banks, M., and Salapatek, P.: Contrast sensitivity From Sensation to Cognition. Vol. I. Basic Visual
function of the infant visual system, Vision Res. Processes, New York, 1975, Academic Press, Inc.,
16:867, 1976. p. 249.
11. Banks, M. S., and Salapatek, P.: Acuity and contrast 21. Miranda, S. B., Hack, M., Fantz, R. L., FanarofF, A.
sensitivity in 1-, 2-, and 3 month-old human infants, A., and Klaus, M. H.: Neonatal pattern vision: A
INVEST. OPHTHALMOL. VISUAL. SCI. 17:361, 1978. predictor of future mental performance? J. Pediatr.
12. Atkinson, J., Braddick, O., and Moar, K.: De- 91:642, 1977.
velopment of contrast sensitivity over the first three 22. Fulton, A. B., Manning, K. A., and Dobson, V.: A
months of life in the human infant, Vision Res. behavioral method for efficient screening of visual
17:1037, 1977. acuity in young infants. II. Clinical application, IN-
13. Atkinson, J., Braddick, O., and Moar, K.: Contrast VEST. OPHTHALMOL. VISUAL SCI. 17:1151, 1978.
sensitivity of the human infant for moving and static 23. Fantz, R. L.: Pattern vision in young infants, Psy-
patterns, Vision Res. 17:1045, 1977. chol. Rec. 8:43, 1958.
14. Allen, J.: Visual acuity development in human in- 24. Teller, D. Y.: The forced-choice preferential looking
fants up to 6 months of age, Thesis, University of procedure: A psychophysical technique for use with
Washington, 1978. human infants, Infant Behavior and Development
15. Allen, J. A., Teller, D. Y., Mayer, D. L., and Lep- (In press.)
per, K.: Development of visual acuity in human in- 25. Atkinson, J., Braddick, O., and Braddick, F.: Acuity
fants up to 6 months of age. (In preparation.) and contrast sensitivity of infant vision, Nature
16. Dobson, V., and Teller, D. Y.: Assessment of visual 247:403, 1974.
acuity in human infants. In Armington, J., Kraus- 26. Teller, D. Y., Allen, J. L., Regal, D. M., and
kopf, J., and Wooten, B., editors: Visual Psy- Mayer, D. L.: Astigmatism and acuity in two pri-
chophysics: Its Physiological Basis, New York, mate infants, INVEST. OPHTHALMOL. VISUAL SCI.
Academic Press, Inc. (In press.) 17:344, 1978.
17. Dobson, V., and Teller, D. Y.: Visual acuity in 27. Gwiazda, J., Brill, S., and Held, R.: Fast measure-
human infants: A review and comparison of be- ment of visual acuity in infants from 2 weeks to 1
havioral and electrophysiological studies, Vision year of age. Presented at Association for Research in
Res. (In press.) Vision and Ophthalmology, Spring Meeting, Sara-
18. Enoch, J. M., and Rabinowicz, I. M.: Early surgery sota, Fla. April-May, 1978.