Sidman & Stoddard 1966 Programming Perception and Learning - Circle Elipse

Programming Perception and Learning
for Retarded Children'

MURRAY SIDMAN
AND
LAWRENCE T. STODDARD
JOSEPH P. KENNEDY, JR. LABORATORIES FOR RESEARCH I N MENTAL RETARDATION,
NEUROLOGY SERVICE, MASSACHUSETTS GENERAL HOSPITAL, BOSTON, MASSACHUSETTS
I. General Methodology ................................ 152

11. T h e Circle-Ellipse Discrimination .....
A. Preliminary Developments ...
C. T h e Second Teaching Program .......
E. T h e Fourth Teaching Program . . . . . . . . . . . . . . . . . . 177
F. T h e Fifth Teaching Program . . . . . . . . . . . . . . 179
IV. T h e Reversal of the Discrimination .................. 197

A. T h e Probe Series ................................ 198
B. T h e Reversal Program .............. 202
References ........................................... 207
This report describes the beginning of a research program that has

three interrelated goals. Our first aim is to develop techniques for evaluat-
ing the behavioral potential of normal and retarded children who are
nonverbal or deficient verbally. Our second aim is to develop techniques
for generating behavior in children who appear to be functioning below
their actual capabilities. Third, we are interested in uncovering basic
processes that are involved when children learn to behave adaptively to
the world around them.
These three lines of work, proceeding simultaneously, constitute a
1 This research was suppoited in part by Public Health Service Research Grant
SB03535 from the Institute of Neurological Disease and' Blindness.
151
152 Murray Sidman and Lawrence T . Stoddard
relatively unusual mixture of basic and applied research (Lindsley, 1956).

Because of the widespread concern with problems of mental retardation
we are attempting here to present our methods and findings in a way that
will be comprehensible to educators, administrators, and teachers who are
interested in applying effective techniques of special education. At the
same time, we describe the methods and data in a form that will permit
our scientific colleagues to evaluate them. We d o not believe our col-
leagues will be seriously hampered by the attempt to share our work
with a wider audience.
At the very outset of our work with retarded children we were struck
by an observation that has formed the basis for all our subsequent in-
vestigations. T h e most pressing initial question is, “How do I communi-
cate with this patient?” Almost without exception investigators and
teachers attempting to answer this question have emphasized the word
“patient” rather than the word “I.” In our work we have taken the
burden from the patient and have ourselves assumed the responsibility
for finding methods of communication, on the assumption that we are
more capable of bearing this responsibility. This assumption has already
begun to pay off. We have found that when we are able to communicate
with the retarded child he is capable of responding, often a t levels
far higher than anyone would have expected.
How does one communicate with a child who understands neither
word nor gesture? T h e problem is difficult enough even with children
who develop normal language skills. T h e formal identity of word sounds
often leads adults to believe that the words they speak mean the same
things when children say them. Experienced clinicians whose job it is
to evaluate children’s behavior know that words d o not equal communi-
cation. Children characteristically interpret the simplest instructions in
ways that are incomprehensible to adult modes of thinking. We now
consider ourselves fortunate in having begun our work with nonverbal
children, for we have avoided the trap of assuming that our verbal in-
structions mean to children what we intend.
1. GENERAL METHODOLOGY
One source from which we may learn much about methods of com-
municating with nonverbal subjects is the large field of behavioral experi-
mentation with subhuman animals. By this we do not equate retarded
children with lower animals, but rather we abstract only one feature
that is common to both. Animals are prime examples of nonverbal orga-
nisms. Nevertheless, animal trainers for centuries and psychologists for
decades have succeeded in “telling” animals what they want them to do.
PROGRAMMING PERCEPTION AND LEARNING 153
It will not degrade the retarded or the immature child if we apply similar
techniques, particularly if we can thereby help him fulfill a behavioral
potential he might not otherwise reach. What has the field of animal
behavioral experimentation to teach us?
A basic feature of much animal experimentation is the artificial en-
vironment in which we observe the animal. T h e experimenter eliminates
potential sources of distraction and uncontrolled influences upon his
subjects, and this permits him to study in a controlled and repeatable
fashion those factors he believes may be significant for the behavioral
development of his subject.
Animal experimentation has also taught us that the hands and eyes
of human experimenters are neither quick nor accurate enough to keep
up with the subject’s behavior, particularly if the experiment is to last
for a significant period of time. One failure to present the subject with
appropriate stimuli at the right time; a slight delay in rewarding him
for desired behavior; a single instance of rewarding him for undesired
behavior: any of these can lead to long lasting or irreparable changes in
the subject’s behavior. A few small errors in recording the behavior can
lead to mistaken conclusions and experiments no one can repeat. For
these reasons we use automatic electronic equipment to program the
experimental procedures and record the data. While such automation
may give a mechanical appearance to a human behavior laboratory,
the equipment is not ordinarily visible to the subject. Its main function
is the yielding of information that is accurate and reliable enough to
form a basis for appraisal and effective remedial action. These considera-
tions, then, have led to the type of controlled and automat,ed experi-
mental environment we shall describe below.
There are two potential disadvantages in using an artificial laboratory
situation to investigate human or animal behavior. When an animal
finds itself removed from familiar surroundings to a strange new environ-
ment it is often frightened-even terror stricken. I t may remain motion-
less for long periods of time, or it may frantically attempt to escape.
Unless the experimenter is particularly interested in the effects of
emotional upset, his conclusions about the subject’s behavioral potential
are likely to be invalid.
Experienced investigators do not begin experimentation until their
animal subjects have become familiar with the new environment and
have learned that it is not dangerous. Strangely enough, those very
experimenters who are so careful about easing the anxieties of their
animal subjects often fail to show the same consideration to human
subjects. Children, and even adults, when they are suddenly placed in a
strange environment, are apprehensive at the very least. We have made
it a practice to help our children adapt to the laboratory situation. To

this end, our laboratory assistants, who generally are attractive, enthusi-
astic, and relaxed young ladies, make friends with the children before
bringing them to the laboratory. They talk to the children, play with
them, give them candy and toys, ride with them in the elevator if such a
trip is required, return with them to their familiar surroundings if they
become apprehensive, and remain with them throughout the experi-
mental sessions. These procedures are not quantifiable, nor can we
describe them objectively. The best we can do is to state that if a child
seems the least bit apprehensive, we try to demonstrate to him (not
FIG. 1. The nine-key matrix.
simply to convince him verbally) that we are to be trusted. Not until

we think we have succeeded do we begin our investigations. Because
we cannot specify the effectiveness of our adaptation procedures we shall
not excuse those instances where our techniques failed by saying that
the child was emotionally upset. Although this was occasionally true,
our data suggest that. for the most part we were successful in handling
the problem.
A second potential disadvantage of an artificial laboratory setting
is that the behavioral findings may also be artificial. They may be quite
unrepresentative of the child's behavior outside the laboratory. This
possibility must be evaluated empirically. The decisive criterion will be
the applicability of our findings outside the laboratory. Meanwhile, we
can only appeal indirectly to existing data that indicate that investigative
techniques similar to ours yield consistent results in widely differing

laboratory situations and, in some cases, even outside the laboratory
(e.g., Krasner & Ullmann, 1965; Ullmann & Krasner, 1965).
All the considerations noted above entered into the design of our
laboratory and our methodology. T h e room we used during our studies
was about 5 feet square, air-conditioned, and painted white and tan,
with sountl-resistant walls and door. Th e child sat on a chair of adjustable
height, facing a square matrix of nine keys, or screens. Figure 1 shows the
nine-key matrix, with the outer keys illuminated and the center key
dark. Figure 2 shows a child pressing a lighted key. On both the left and
FIG.2. A child seated in front of the nine-key matrix.
right sides are plastic trays into which automatic devices deliver such
rewards as candies, toys, or tokens that the child can trade later for
other things he may desire. Just below the keys is a shelf on which
the subject may place his rewards. An additional device, not shown in
Figs. 1 and 2, is a set of chimes high on the wall above the child’s chair.
Each of the nine individual keys in the matrix is a translucent plastic
(Polacoat) 2 x P-inch square. Stimuli-squares, circles, words, etc.-can
be projected on the keys from the rear. We used a Leitz automatic slide
projector located on the other side of the wall from the subject. When
the stimuli were displayed to the child his task was to press one of the
keys. A small microswitch behind each key signaled the child’s choice
to our automatic control and recording apparatus. Photocells beside
the keys indicated to the control and recording equipment the key that
156 Miirray Sidrnan and Lnwrence T.Stoddard
was correct on any given trial. A11 the stimuli to be presented to the
child on a given trial, along with the pattern of illumination of the
photocells, were photographed on 35 mm color film and mounted in
slides. .A shutter on the projector kept transient light from the photo-
cells while a slide was being changed. Shutters were also mounted behind
the keys so that stimuli could be made to appear and disappear rapidly.
One and one-half seconds elapsed between the end of one trial and the
beginning of the next trial. T h e motor-driven shutters, the slide pro-
jector, the photocells, the rewarding devices, and the recording equip-
ment were controlled automatically by electronic components. [Hively
(1964a) and Holland (1961b) have described similar apparatus.]
When the child pressed the correct key, the chimes sounded and a
reward dispenser operated. Most children received small, candy-coated
chocolates (M & M’s) as their reward. Factors such as diet restrictions or
an occasional child’s unwillingness to eat candy sometimes required
us to use other kinds of food, or tokens that the children could trade
for toys. We used whatever reward kept the child at the task we were
trying to teach him. Some of our procedures also demanded that a
tangible reward not be delivered to the child every time he pressed the
correct key. On those instances we rewarded him for only a certain
percentage of his correct choices (variable-ratio schedule), but the chimes
always rang after a correct choice even when the child received no other
reward. T h e same key was never correct on consecutive trials,
T h e first task we set ourselves was the development of a simple tech-
nique for evaluating visual perception i n severely retarded, nonverbal
patients. T o do this we selected two simple geometric forms, the circle
and the ellipse; the child’s task was to discriminate between them. These
forms have at least two features that make them particularly useful.
If we start with a very flat ellipse and gradually increase its vertical
dimension (the minor axis), the ellipse gradually becomes more and more
like a circle. I n fact, the circle may be considered simply as an extreme
form of ellipse, with equal minor and major axes. Normal adults can,
with little difficulty, distinguish a circle from a n ellipse whose minor
axis is nine-tenths as long as its major axis. This provides us with a
quantitative measure of a person’s ability to distinguish small differences
in his visual surroundings. Technically, the smallest difference he can
perceive is called the “difference threshold.”
Besides being easily adaptable to quantitative measurement, circles
and ellipses have the advantage of being easily prepared for presentation
to a subject. They require little more than a steady hand to operate a n
ellipsograph, plus a bit of skill and attention to detail in photographing
PKOGKAMMING PERCEPTION A N D LEARNING 157
and mounting them in slides for projection. They need no complicated

or expensive optical equipment.
11. THE CIRCLE-ELLIPSE DISCRIMINATION
How do you go about getting a nonverbal, severely retarded child to

tell you if he can distinguish circles from ellipses? T h e best way to answer
this question is to describe what we did and the results we achieved with
our first patient. Actually, our extensive work with this patient has been
so interesting that we shall publish his story in detail elsewhere; how-
ever, we may describe him briefly here.
A. Preliminary Developments
Our first patient, (C.M.), classified as a microcephalic idiot, was 40 years
old and had resided in an institution for the retarded since he was 6.
His IQ had never been rated higher than 20, and this was before he was
10 years old. Since that time it had steadily declined. Although C.M.
was not a child in chronological age, his behavior was so impoverished
that he served as a standard for evaluating the applicability of our
techniques to the most severely retarded persons whom we could hope
to reach. When we first saw C.M., he spoke only three words, “hi,” “bye,”
and “yah.” He often used even these few words inappropriately. His
motor coordination was good, and he did respond to simple commands
such as “Come here,” and “Take off your coat” (which he accomplished
only with some help), and sometimes he seemed to respond to his own
name. He could feed himself without utensils, was only partially toilet
trained, and displayed little or no social behavior, spending most of his
time in front of a television set, attending to it whether the picture was
clear or completely blurred. H e did have a certain amount of imitative
behavior; he would touch an object, lift it, or drop it, repeating these
actions after someone else. I t was this minimal imitative repertoire of
behavior with which we started our work.
Preliminary testing had suggested to us that C.M. could not distin-
guish a circle from an ellipse, but his general understanding was so
limited that we suspected his comprehension of what we were asking him
to do. We therefore planned a series of steps by means of which we hoped
to achieve the necessary communication.
We first loaded our automatic dispenser with M & M candies that we
had previously ascertained he would eat. We then sat C.M. in front of
our matrix of keys and delivered a candy into the plastic tray. T h e
sound attracted his attention and he picked up and ate the candy. We did
this several more times and C.M. learned easily to reach into the tray
and take the candy immediately after it dropped. We then turned on
the slide projector, which projected a bright light onto one key, the
others remaining dark. We pressed the key once ourselves, the chimes
rang, and a candy was immediately delivered; we then took his finger
and pressed it against the lighted key-a different key this time, since
the projector had automatically advanced to the next slide-and again
a candy was delivered. After three such demonstrations C.M. pressed a
key himself, and from that point he was on his own. H e received addi-
tional help from only two sources: a candy, which was a consequence
of pressing the correct key, and the methods by which we presented stim-
uli to him.
C.M. learned with little difficulty to press only the one key that was
lighted on any given trial and to reject dark keys. We then projected
forms-circles, squares, and X’s-o n to the lighted key. He did not have
to observe these forms; he could still procure the candy simply by press-
ing the lighted key. To draw his attention to the forms, we gradually
increased the brightness of the incorrect keys, without, however, project-
ing any forms on these incorrect keys. At the end of this process, which
we shall describe more completely later, all keys were equally bright
but only the correct key had a circle, square, o r X superimposed on the
bright background. C.M. had learned to select not simply the bright
key but the key that had a form projected on it.
We had thus established some elementary but necessary behavior.
C.M. still did not have to observe what the forms were; he had only to
tell the difference between a bright key with a form on it-any form-
and a bright key without a form. Having taught him this preliminary
behavior, we were now in a position to move toward our goal of finding
out how fine a discrimination he could make between a circle and an
ellipse. T h e next step was to teach him to tell the difference between a
circle and a very flat ellipse. We certainly could not test the fineness of
his perception until we had established whether he could make this
elementary distinction. In fact, as our subsequent. data will show, if we
had not deliberately taught him to tell circles from ellipses, we would
have been led to the entirely erroneous conclusion that he was incapable
of distinguishing them.
Instead of simply testing his discriminative capacity, we continued
along the lines we had already begun; we designed a program to teach
C.M. to tell the difference between circles and ellipses. This introduces
a principle that underlies all our work and, we believe, must form the
basis for all techniques of behavioral evaluation whether the subjects be
retarded or normal. T h e principle, expressed simply, is, “Don’t test;
1’KOGRAM.IIIKG PERCEPTION A S D LEARNING 159
teach.” Tests tell us what a person has already learned; they may say
little about what a person is capable of learning. We must try to teach
a person the performance we want to evaluate, using our ingenuity to
increase the effectiveness of available teaching techniques. Otherwise we
shall test not his adequacy but rather the adequacy of his teachers and
of those aspects of the environment which we normally allow by default
to accomplish the task of teaching our young.
B. The First Teaching Program
\Ye prepared a series of 72 slides as follows. On each of the first few
slides only one of the keys was bright, and the others were dark. T h e
position of the correct key always changed from trial to trial. In addition,
the bright key also had a circle projected on it (Fig. 2). This was some-
thing C.M. had already learned; he pressed the lighted key and rejected
the dark ke)s. But the circle was simply gratuitous; he did not have to
A B C D
FIG. 3. Schcmatic illustration of widely spaced steps in the fading process. T h e
correct key always had the circle on a bright background. The incorrect keys were
dark at first (A), and gradually became brighter (B, C, I)). As the incorrect keys became
brighter, the contrast between the ellipses and the background also increased.
observe that anything was on the lighted key, let alone that it was a
circle. Then, on successive slides, we very gradually increased the bright-
ness of the incorrect keys. So far, the process was no different from those
he had already experienced. T h e new feature was the way we introduced
the ellipse. As the incorrect keys gradually became brighter, a faint, very
flat ellipse appeared upon them. As the keys continued to brighten on
successive trials, the ellipses also became more and more distinct. As
we gradually faded in the bright background on the incorrect keys, we
simultaneously faded in the flat ellipse. Figure 3 illustrates the process
schematically. Note that the center key remained dark. This was true
throughout the work to be reported here and we shall refer to the matrix
as if it had only eight keys. T h e center key was being reserved for subse-
quen t matching-to-sample procedures. T h e circles and ellipses were black
forms. T h e background, when fully faded in, was bright yellow. Yellow
was used because it was cheerfully colorful, provided good contrast with
the black forms, and had less glare than a white background.
1GO Murray Sidman and Lawrence T . Stoddard
We started, then, by building on behavior we had already taught

C.M., namely, to discriminate between bright and dark keys. Then, in
very small steps, we gradually transformed the basis for his discrimina-
tion to “key-with-circleversus keys-with-ellipse.”At the beginning of this
program C.M. was able to base his selection on brightness alone; then,
we gradually changed the basis for his discrimination so that he could
select the key with a distinct figure on it out of several keys with rela-
tively indistinct figures and with dimmer backgrounds; finally these
criteria were gradually faded out until the only remaining basis for
choice was the actual shape of the forms, i.e., the circle and the ellipse.
By changing the criteria in very small steps, we expected to help our
subject understand what was being asked of him. T h e fading program
was our way of telling C.M. what we wanted him to do; the candy was
our way of telling him he was doing it correctly.
How did we tell C.M. about his errors? This is a more complex ques-
tion than it might appear, and our method of dealing with errors
deserves some discussion at this point. At the same time, it will also be
expedient to describe one of our techniques for recording the subject’s
behavior. But before doing this we may relieve any feelings of suspense
in the reader by stating that C.M. did learn to discriminate the circle
from the flat ellipse. Although he made some errors in the process, his
severely impoverished behavior made us quite certain that he could
not have learned it at all without the kinds of help we gave him. Let
us look first at our method of recording the subject’s behavior and then,
in more detail, at his performance.
Figure 4A shows a picture of the projector and the kymograph-type
recorder. Figure 4B is a closeup view of the recorder. An electric motor
drives the paper continuously in the direction indicated by the arrows, a.
As the paper moves, it passes under a pen that traces a continuous line.
The pen itself is attached to a chain and pulley; the chain, in turn, is
attached to the rear of the projector’s slide tray. As the slide tray advances
to project successive slides, it pulls the pen up the paper in the direction
of the arrows, b. Each time the subject presses the correct key, the slide
tray advances to project the next slide and draws the pen one step along
the paper. During the time between the subject’s responses, the pen
simply draws a straight line along the paper.2 At h in Fig. 4B, the pen
records the consequences of pressing the wrong key; this feature will be
described more fully later.
Figure 5 shows the record of a brighter-than-average 4-year-old girl.
Each step upward indicates that this little girl pressed the correct key.
2 This recording technique was first used by James G . Holland.
FIG.4. A, The projector and recorder; B, a closeup of the recorder. The arrows, a,
indicate the direction in which the paper moves at a constant speed. The arrows, b,
show the direction in which the slide tray pulls the pen. c, Recorder pen; d , slide
tray; e, chain and pulley connecting the slide tray to the recorder pen; f, extra pens
for marking rewards and other events; g, steps drawn by recorder pen: h , consequence8
of an error.
162 Murray Sidnzan and Lawrejice T . Stoddard
T h e horizontal distance between steps shows how much time elapsed

between successive choices. T h e time scale appears below the record
and the numbers on the left indicate that there was a total of 80 slides
in this particular series. T h e numbered dots above some of the steps
indicate slides on which the background brightness of the incorrect keys
(on which the ellipse also gradually appeared) was increased. There were
20 brightness increments in this fading process as the incorrect keys
changed from very dark to a brightness equal to the correct key-the
80- 1 '
75.
70.
65.
60.
m55.
W 7
osO.
0 4 sy ~ ~ 4
,>I*
540-
I-35.
E30.
5 5 .
82c.
IS.
one with the circle on it. T h e actual brightness values are almost im-
possible to specify, but it is evident that 20 increments from darkness
to full brightness represent a relatively slow process.3
After slide 40 it was necessary to insert a new slide tray into the
projector; the line and arrows indicate the point at which we did this.
3 In the photographic piocess, Bourges overlays of varying degrees of opacity were

placed over the incorrect stimuli. The degree of opacity was reduced to permit the
brightness and contrast of the incorrect keys to increase. T h e process was thoroughly
empirical. The actual brightness values are a function of such factors as: the brightness
of the paper that formed the background; the type, intensity, and distance of the
lights used in photography; the type of slide projector and projector bulb; the distance
of the projcctor from the keys; the illumination in the experimental room; etc.
FROGRAMMING PERCEPTION AND LEARNING 163
T o minimize any disruptive effects of the interruption, the first few

slides in the second tray reviewed brightness increments 14 and 15 before
going on to the rest of the sequence.
T h e record in Fig. 5 actually represents the second stage of the circle-
ellipse program, but it serves to illustrate a perfect performance. This is
the type of record toward which all of our efforts were directed, even
with severely retarded children and with normal children no older
than 3 years. As our efforts approached success, the records came to
-
ONE MINUTE
FIG. 6. T h e performance of a severely retarded subject o n the first circle-ellipse
program. Th e numbered dots indicate the last slide of each successive brightness
increment. a indicates the subject’s first error, followed by the reversal of the slide
tray after he corrected his error. Each of the small marks, like the one at a, indicates
an incorrect choice. T h e arrow shows the end of the first slide tray. Although C.M.
made three errors on the first slide of the new tray (after the arrow), there was no
backup from the first slide af the tray. An apparatus failure prevented the backup on
one occasion in fading increment 14.
represent to us the actual progress of the child, moving upward step by

step toward new achievement. One’s feeling for these upward steps as
the actual climb of the child may be intensified by examining a record
that shows the subject making errors on his way up.
Figure G is the performance of our first subject, C.M., microcephalic
and severely retarded. T h e figure also illustrates our procedure for han-
dling errors. T h e numbered dots indicate the last slide in each successive
increment of the fading sequence. For example, the first brightness level,
in which the wrong keys were completely dark, contained four slides.
The record has been cut into short segments and rearranged for more
compact presentation; as in the previous figure, the progression is from
bottom to top. If we look at the three segments at the bottom, we see
that C.M. made no errors through the first six brightness levels. He
worked very slowly during the first four brightness levels and then
speeded up considerably during increments 5 and 6, as indicated by the
shorter distance between steps in the third segment from the bottom.
C.M. made his first error on slide 24, and the error is indicated by
the small oblique mark on the record (identified by the letter, a ) , When
he pressed the wrong key, nothing happened; the slide did not change
and the stimuli remained exactly as they were. T h e next key he pressed
was the circle, but then, instead of advancing to the next slide, the pro-
jector reversed and repeated slide 23. This is our standard procedure:
when the subject presses a wrong key, the stimuli remain until he has
corrected his error; when he finally presses the circle, the slide tray
reverses and he is again exposed to the preceding array of stimuli. T h e
slide tray advances only if the subject’s first choice on a given trial is
correct. (See also Fig. 4B, at h.)
An error, then, represents a step backward and shows up as such on the
record.4 In spite of the anguish this causes us whenever it happens, the
“backup” procedure has three advantages. First, if the subject comes to a
difficult part of the program, his errors automatically return him to a
less demanding level. Instead of leaving him to flounder, we gave him
this extra help (Hively, 1964b).
T h e second advantage of the backup procedure is that it helps us
understand whether the subject’s error resulted from some inadequacy
in the fading sequence-we did not adequately communicate what we
wanted him to do--or whether the source of the error was relatively
trivial, such as a temporary lapse of attention. If the error arose from
unimportant sources, the subject will choose the correct key when that
particular slide reappears. For example, after C.M. pressed a wrong key
on slide 24, the slide tray reversed to slide 23; he made a second correct
choice on that slide and when he was again presented with slide 24,
he chose the correct key. Similar errors, which C.M. did not repeat
after the backup, also appear on slides 30, 34, 35, and 38.
On slide 46, however, we see a different kind of error. Our subject
had climbed without error through slides 42, 43, 44, and 45, but on slide
46 he made two wrong choices before finally choosing the circle. The
4 The authors are indebted to James G. Holland for the backup procedure, which
has become a standard feature of our programming techniques.
PROGKAMMING PERCEPTION AND LEARNING 165
tray reversed to slide 45 and again he made two errors, even though
he had initially chosen the circle on his previous exposure to that slide.
When he went back to slide 44 his first choice was again an error, and
on slide 43 he made six errors before picking the circle. He was finally
correct on slide 42, made two errors again on slide 43, went back again
to slide 42, and again climbed successfully to slide 46.
We have described here a most common and revealing phenomenon.
It may be summarized by stating, “Errors create more errors.” [See
Terrace (1963a,b) for the first unequivocal demonstration of the effective-
ness of fading techniques in producing errorless learning.] After meeting
a difficult slide, the subject backed down through the program, making
errors where he had previously had no trouble at all. T h e reason for
our anguish at this point was not simply that our subject slid backward,
but that we were to blame for his difficulties. Our program had failed to
communicate adequately to him what he was to do; perhaps our fading
series was progressing too rapidly for him to follow our instructions.
When he made his first error on slide 46, he was presumably using the
criteria for selection that had been successful up to that point. However,
the error showed these criteria to be wrong. Therefore, when slide 45
appeared again, he changed his basis for choice, and where he had
previously been correct he was now wrong. He then continued to have
trouble throughout brightness increment 13, oscillating up and down,
occasionally recovering his performance but always meeting the same
troublesome obstruction on slide 46. Eventually he passed through this
difficult region but continued to have trouble when we increased the
brightness of the incorrect keys a bit more (increment 14), and he made
little progress through 15 (before the 1-hour rest).
The processes involved here are thoroughly normal. Whoever our
subject may be, retardate, normal student, or psychotic patient, if he
finds himself inadequately instructed he will oscillate between alterna-
tive modes of action. Unless he happens by chance to hit upon a correct
solution, he will show no further progress.
When the backup procedure shows the subject making errors where
he had previously been correct, we know where our program is inade-
quate and where we must revise it. A large part of our story is, in fact,
a description of the successive revisions through which our program
passed. The results of those revisions will substantiate our contention
that inadequate performance is more likely to result from deficiencies
in our techniques of evaluation than from deficiencies in the children
(Bijou, 1964; Holland, 1961a).
The third advantage of the backup procedure stems from its contribu-
tion to our basic understanding of the learning process. The traditional
practice of drawing learning curves loses its meaning when we revise

o ur program and the subject then passes beyond a level at which his
forward progress had previously halted (Hively, 1962; Moore & Goldia-
mond, 1964).
We interrupted our first experiment and gave C.M. a rest for
1 hour. We then began again with the incorrect keys at brightness level
12, through which C.M. had previously passed relatively easily. This time
he hatl more trouble and had even more difficulty at level 13 than he
had the first time through. It is possible that the break in continuity made
it difficult for him to recover his earlier performance. After oscillating
for some time between slides 46 and 41, C.M. finally seemed to hit upon
a n effective procedure. H e then passed through the remaining levels of
the program with little difficulty, and actually made n o errors in the
last two brightness increments. In the seven final slides the brightness
cue hatl completely disappeared and C.M. consistently selected circles
and rejected ellipses.
M'e had mixed feelings about C.M.'s eventual success. We were pleased
for him, and excited because it revealed a level of performance that was
previously unsuspected. It also put us in a position to achieve our imme-
diate goal, which was to assess his perceptual capacity by finding out
how much we could make the ellipse look like a circle and still have him
discriminate the two forms. But we were not at all satisfied with the
number of errors our teaching program had created. If we were to make
general use of this technique for perceptual evaluation, we would have
to increase its effectiveness considerably.
Can we evaluate our teaching program quantitatively? There are a
numher of ways we might try to do this. For example, if we combine
both parts of the session, before and after the rest period, we find that
C.M. had a total of 183 trials. Of these, his first response was correct
on 129, or 707". We might separate two types of errors, those which
occurred the first time a slide was presented and those which occurred
when a slide reappeared after an error. These errors arise from different
sources antl it is legitimate to consider only the errors on a subject's
first exposure to a slide as being a product of program inadequacies;
subsequent errors are likely to be the result of preceding errors. 1,ooking
at the session this way, we see that C.M. was exposed to 55 slides before
the rest period antl 32 slides later, a total of 8 i . Of these S i , he was
correct on 65, or i4%, the first time he was exposed to them.
IVe could perhaps make a more accurate evaluation by taking into
account the total number of errors the subject made o n each trial. Only
one key was correct antl seven were incorrect per trial. If the subject
were simply pressing keys at random, we could expect him to pick the
correct key on one out of eight trials, or 121/2y0. If he eliminated a

wrong key from consideration after pressing it, he would have only one
chance in seven of picking the correct key on his second choice, one
chance in six on his third choice, and so on. Therefore, when C.M. made
only one or two errors on a trial, he still showed a much better per-
formance than purely random selection would have produced. A single
error indicates that the subject is far from being completely bewildered
and that the program still carries considerable effectiveness. But when
the subject makes several errors on a given trial, there can be no doubt
that the behavioral control exerted by the program has broken down
and needs repair.
We did not feel it worthwhile at this point to carry out a sophisticated
statistical analysis of C.M.’s performance. We were satisfied-or, more
accurately, dissatisfied-that we did not have anything resembling a n
adequate teaching program. Even though this particular subject was
able to go on to more difficult perceptual tasks, we wanted a teaching
instrument that was likely to succeed reliably with a large population
of severely retarded subjects. Therefore, after taking a close look at the
areas in which our first subject experienced difficulty, we revised the
program.
C. The Second Teaching Program

C.M. made his first error at brightness level 7, but this was not repeated
and not followed by any wrong choices in level 8. However, he did make
errors on the initial slide of increment 9, and then made at least one
at levels 10, 11, and 12, finally exlperiencing great difficulty in levels
13-15. C.M.’s performance indicated that the early stages of the fading
program were more than adequate, and so our first change was to shorten
this portion of the program. Keeping the same number of brightness
levels, we reduced the number of slides to two in levels 1 through 8.
Since C.M.’s difficulties seemed to begin with increment 9, we inserted
an intermediate brightness level between the former increments 8 and 9.
These changes can more easily be followed by referring to Fig. 7. We see
here that brightness increments (fading steps) 1-8 are identical in pro-
grams 1 and 2, but that the number of slides per increment has been
reduced to two in the revised program. Level 9, a new fading increment
(2 slides), now follows the old level 8, and the former level 9 now becomes
10, with four instead of three slides. Level 11 of the revised program is
also a new facling increment, intermediate between former levels 9
and 10. Levels 12 and 13, corresponding to the old levels 10 and 11,
remain unchanged.
Since the slide tray of our automatic projector held only 40 slides,
168 Murray Sidman nnd Lawrence T . Stoddard
we had to stop briefly after slide 40 to change trays. In the first program
this change occurred after increment 11, Since we had reduced the num-
ber of slides in the early stages of the revised program, we were able to
proceed to level 15 before having to stop and change slide trays. At this
point we made another revision of the program. Reference to Fig. 6
shows multiple errors on slide 41, suggesting that the interruption of
the procedure caused C.M. some difficu1t.y. In the revision, therefore,
we backtracked at the beginning of the second slide tray and presented
CONSECUTIVE SLIDES
FIG.7. Schematic illustration of the way the brightness increments (fading steps)
changed on successive slides of the first four circle-ellipse programs. The programs are
numbered consecutively from 1 to 4. The fading steps do not represent equal brightness
increments.
additional slides of levels 14 and 15-levels the subject must have com-
pleted successfully to reach the second tray. This also had the effect of
increasing the number of slides at these levels, giving the subject six
opportunities at level 14 and seven at 15 (former levels 12 and 13).
At this point in our thinking, we were still tied to the notion that
mere repetition (with reward) would help the subject pass through a
given step of the program. Our later work caused us to discard this time-
honored notion but it was responsible, in the first revision of the pro-
gram, for adding more slides in the final stages. We added four slides
to level 16 (formerly 14), two slides to levels 17 and 18, and one slide to
level 19. The final brightness increment, formerly 18 and now 20,
contained the same number of slides as before. At this level, of course,
the brightness difference between correct and incorrect keys had dis-
appeared and the subject had to make his selection on the basis of circle
versus ellipse, the performance toward which the program was directed.
The revisions seemed relatively reasonable. The initial part of the
program, in which the subject could simply discriminate the bright key
from the darker ones, became shorter; more steps were added to the
intermediate parts of the program, where the basis for the subject’s
FIG. 8. The performance of a normal child, age 3 years and 8 months, on the second
circle-ellipse program. The arrow indicates the change of slide trays. See text for
explanation of letters. The numbered dots indicate the end of each consecutive bright-
ness increment.
choice was beginning to shift from the bright-dark criterion to the circle-
versus-ellipse criterion. We added additional slides near the end of the
program where the brightness cue had all but disappeared in order to
give the subject more practice at each level.
Figure 5, the record of subject N.S., a normal 4 year old, was obtained
with the revised program. On this record, levels 9 and 11 are the new
ones, the repetition of levels 14 and 15 is indicated at the beginning
of the second slide tray, and the number of slides a t each step may
easily be seen.
Figure 8 shows the record of another normal child, P.C., 3 years and
8 months old. This child made no errors from level 1 through 15, the
first 40 slides of the revised program. His first error (at a on the record)
came in the second slide tray, on the repeated presentation of step 15,
but this error caused him no subsequent difficulty. His second error
came at b, the first slide of level 17. T h e subsequent return to level 16

helped the child get over this hurdle, and he progressed relatively rapidly
through the remainder of level 17 with one more error (at c ) . His next
wrong choice (at d ) also came on the first slide of a new brightness
increment and again he was helped by returning to the previous level.
P.C. made errors on six trials of step 18. Only at e , however, did he
press more than one key per trial, indicating a relatively high per-
formance in spite of the frequent errors. Two of the errors, at e and g,
were on slide 66, and the errors at h and i were both on slide 68. T h e
child then went through level 19 with only one error, at j , but had con-
siderable trouble with slide 75, the second slide of level 20, the criterion
performance. T h e third error on slide 75 (at rn) also led to an error on
slide 74, on which the child had previously made three correct choices.
T h e error on slide 74 (at n) caused the slide tray t o reverse back to level
19 and from this point on, P.C. completed the program with no more
wrong choices.
Level 18, on which P.C. experienced his first real difficulty, corresponds
to level 16 in the original program. By the time C.M., our first subject,
had reached level 16 he had overcome his major troubles with the
program (Fig. 6). P.C., however, climbed rapidly through the early and
intermediate steps, experiencing difficulty only when he was relatively
near the end. At level 18 the brightness cue had become extremely diffi-
cult, if not impossible, to use, and we suspected that new factors were
responsible for P.C.’s troubles at this stage. We had anticipated a possi-
ble satiation effect, and starting at the beginning of the second slide
tray (after slide 40) we gave P.C. a candy for only 50% of his correct
choices, in an irregular order. This did cut his candy consumption in
half but may not have prevented satiation. Altogether, he received a
total of 74 candies, 54 by the end of level 17. We shall have more to
say about the problem of satiation later.
Figure 9 shows a particularly interesting record. L.C. was a 5-year-old
child who was diagnosed as autistic (Kanner, 1943; Kanner, 1944). Autism
is by no means a well-defined disease, and it probably encompasses a
number of different types of behavioral disturbance. An outstanding
feature of autistic children is their lack of constructive interaction with
their environment. Such children range from those who are almost
completely unresponsive and inactive to those who are so highly over-
active that they are a menace to their surroundings. L.C. fluctuated
between these extremes. It was all but impossible to communicate with
him, and we had no way of knowing whether he was actually incapable
of constructive behavior or whether his failure to display such behavior
was a function of our inability to make contact with him. On our circle-
ellipse program, L.C.’s performance was somewhat disappointing, since
he did not reach the final level of discrimination, but he did display
more constructive behavior than we had seen from him up to this point.
His record (Fig. 9) is remarkably like that of C.M. (Fig. 6). L.C. experi-
enced his greatest trouble when he first met brightness increment 15,
which corresponds to 13 for C.M. L.C., however, did succeed in passing
through level 15 at the end of the first slide tray. Reexposing him to
’ONE MINUTE’
FIG. 9. The performance of an autistic 5 year old on the second circle-ellipse
program.
levels 14 and 15 at the beginning of the second slide tray seemed to help,
but he finally stalled completely at level 17. Unlike C.M., however, L.C.
worked quite rapidly and relatively effectively up to the point where
he first experienced great difficulty. With the reduced number of slides
in the early part of the revised program, L.C. reached level 15 much more
rapidly than C.M. had done.
I n order to forest,all satiation, the intermittent schedule of reward was
introduced for L.C. at slide 14 (the end of level 7). The failure to receive
candy on some trials did not disturb him at this point, as his rapid
progress through the next few steps indicated.
Nonetheless, beginning with the last slide of level 13, L.C.’s behavior
changed markedly. At this point, after having received 29 candies, L.C.
seemed to lose interest in what he was doing. During the frequent long
pauses after level 13, the child usually left his chair and played else-
172 Murray Sidnzan and Lawrence T . Stoddard
where in the room, occasionally returning to the keys. The frequent

occurrence of multiple errors on a trial indicates that even when he
returned to the keys he seemed to press indiscriminately without paying
much attention to what was on them. As in the case of P.C., we suspected
some such factor as satiation or boredom. Encouraging as his performance
was, it still did not go far enough. We were unable to establish the
circle-ellipse discrimination that we needed to test his discriminative
capacity.
Before going on to describe our next revision of the program, we must
present one more child, for he gave us the courage to make a radical revi-
sion. R.L. was another autistic child, older (8 years, 5 months) than L.C.,
but with much more severe symptoms. This child was so hyperactive
that he had to be restrained to prevent him from destroying everything
about him. He rarely spoke, and it had not been possible to engage his
attention for more than a few seconds at a time. This seeming inability
to hold still, together with the general inappropriateness of his responses
to the social and physical environment, made it impossible to evaluate
his behavioral capacity by any conventional means.
When R.L. was brought to our laboratory and released from his
physical restraint, he immediately began to dart about rapidly, without
any apparent purpose to his movements. When he was offered a candy
he knocked it from the experimenter’s hand and disregarded it. He then
darted out the door, with the experimenter close behind. For the next
10 or 15 minutes the two of them, child and experimenter, ran together
through the halls of the building, passing in and out of every door along
the way, running upstairs and downstairs, never pausing for more than
a few seconds in any one place. After a while the child became trusting
enough to hold the experimenter’s hand during their travels, and it
was possible gradually to lead him back to the laboratory. After passing
through the door of the experimental room, the experimenter pressed
the lighted key of slide 1, the chimes rang, and a candy fell noisily into
the tray. The child immediately seized and ate the candy, sat himself
down in front of the keys and went to work. His record appears in Fig. 10.
R.L.’s performance was quite remarkable. He went through the pro-
gram rapidly and accurately, making errors on only two slides-the
first slide of level 17 and the third slide of level 19. At the end of about
25 minutes he had learned the circle-ellipse discrimination and, in fact,
went on to other programs, working steadily for over an hour.
We had taken the precaution of depriving R.L. of breakfast before
bringing him to the laboratory. This undoubtedly contributed to the
large number of candies, 82, he was willing to earn and eat.
A more extended report on R.L. will be prepared later, in conjunction

with a longer series of similar children. At this point in our work,
however, he demonstrated to us that we were on our way to developing
a technique we could use both to teach supposedly unteachable children
and to investigate the behavior of children with whom we could not
communicate by ordinary means.
We tried the first revision of our circle-ellipse program with seven
children. Of these seven, only three reached the criterion performance,
FIG. 10.. The performance of an autistic child, age 8 years and 5 months, on the
second circle-ellipse program. The record of trials 1-4 was lost through apparatus
failure.
learning to discriminate the circle from the flat ellipses. One of the
children who went all the way was a normal 4 year old who made no
errors (Fig. 5); the second was a normal child, age 3 years and 8 months,
who made errors on 9 slides the first time he came to them (Fig. 8).
These first-time errors are the ones that concern us, for the errors a
child makes on reversing back to a slide he had previously done correctly
are probably a direct result of the earlier errors. The third child who
went all the way through the program was an autistic child (Fig. lo),
who made only two errors. One autistic 5 year old (Fig. 9) went through
the program only to level 17; another normal child, age 3 years and
8 months, went through the program only to level 14. Our youngest
subject, age 1 year and 11 months, went through only seven slides of
174 Murray Sidman a n d Lawrence T . Stoddard
the program, with three errors, before stopping. The seventh child, a
severely retarded 4% year old, did not even get started.
This summary demonstrates that the first revision of our program
still did not provide the instrument we needed. On the other hand, the
results were encouraging enough in a few cases to make it appear worth-
while to continue our efforts. In the interest of making the program
easier for the children, we had lengthened it in our first revision. T h e
data made it look as though this was a mistake. Although the insertion
of new steps did seem to produce relatively smooth performances through
the first slide tray, a number of the children accumulated errors in the
last part of the program. These errors at the end might have come from
several sources. For example, the fading increments at the end of the
program might have been too large, or the children might have become
satiated or bored. Our first data (Fig. 6) suggested that once a child
passed through the difficult intermediate steps of the program he would
then go through the final few steps relatively easily; he would have
learned the circle-ellipse discrimination even before the brightness cue
had completely faded out. This was not true, however, for any of the
other children except R.L. and N.S. T h e rest either stopped working
before they reached the end of the program or made a large number of
errors on the final few series of slides.
A major indication that the source of the trouble was the decreasing
effectiveness of the candy rather than the program itself comes from
the fact that in the final few steps the children usually made errors not
on the first slide of a new brightness increment but on subsequent slides,
after they had already made some correct choices. If the increment were
too large, we would expect the children to make an error on their first
exposure to a new level.
Our success with R.L., whom we had deliberately made hungry, was
another bit of evidence that satiation was partly responsible for the
children’s errors in the late stages of the program. Hively (1962) has
also presented data leading to a similar conclusion.
Admittedly, our evidence was inconclusive, and later we discovered
additional explanations for the program’s inadequacy, but these were
our reasons for the next step we took. In addition to adding a few inter-
mediate steps to the program, we reduced its length drastically.
D. The Third Teaching Program
A glance at Fig. 7 will show how the consecutive slides and fading
increments of program 3 compare with the earlier versions. Program 3
had only one slide per brightness level, and the children reached the
end in only 24 slides. This is approximately one-third the number of
slides required in the first two versions of the program, Although several
intermediate brightness levels were inserted in the later portion of pro-
gram 3, some of the early levels were removed, so that program 3 had
only one more fading step (21) than program 2. We hoped to give the
children help where their errors showed they needed it.
The youngest child who climbed all the way through program 3 was
E.F.C., a normal 3 year old (Fig. 11). He made some errors at the very
beginning while he was finding out, what we were asking of him, and
I
I
’ONE MINUTE
FIG. 11. The performance of a normal 3 year old on the third circle-ellipse program.
then went through the rest of the program with errors on only one slide.
The brightness increments are no longer numbered on the record
because there is only one slide per brightness level. The last three slides
represent the criterion performance, in which all keys were equally
bright.
S.A.C., age 2% was the youngest child who tried program 3, and the
program failed to carry him all the way through. His records on two
different days appear in Fig. 12. The lower record of the figure shows
S.A.C. “hanging up” badly at about brightness level 15, and finally his
performance deteriorated almost completely. On his second try, he
again reached level 15 but could proceed no further. The consistency
I-
+ SAC.
c~~~ MINUTE
FIG. 12. Two records of a normal child, age 2& years, to whom the third program
failed to teach the circle-ellipse discrimination.
176 Murray Sidirian orid Lnwrence T . Stoddard
of the data in these two sessions illustrates that the mere presence of a
circle on the correct key was no guarantee that the child would observe
and use the circle as his basis for choice. As long as the correct key was
distinctly brighter than the incorrect keys, S.A.C. followed our “instruc-
tions’’ quite successfully. However, when the brightness cue began to
fail him, he did not transfer the basis for his choice from brightness
to the presence of the circle.
.Among the children who tried program 3 was one hyperactive autistic
girl, age 3 years and 3 months, who was described as “organically driven”
byethe examining neurologist and was rated “probably of average intelli-
gence” by the clinical psychologist who administered standard psycho-
logical tests. M.B.’s record appears in Fig. 13. She made errors on several
slides, but until she reached level 16 the backup procedure seemed to
I ,-A J
,
~~~~
~ --- U L , % -
-,_r
-r?L- --
, r7 -2--
b/I/bI
ONE MINUTE
FIG. 13. The performance of an autistic child, age 3 years and 3 months, on the
third circle-ellipse program.
provide her with the extra help she needed. In spite of a long period
of oscillation in the neighborhood of level 15, M.B. finally broke through
and had little trouble with the final few slides of the program. More
remarkable than the successful completion of the fading program was
her persistence even through the levels she found difficult. It was not
ordinarily possible to keep this child’s attention on any one aspect of
her environment except through the most intense effort. In this respect
she resembles the other autistic children who tried our program, and we
wonder what heights they might reach if we carefully programmed their
total environment for 24 hours each day instead of just the few minut,es
our small program required (Ferster, 1961; Ferster & DeMeyer, 1961;
Ferster & DeMeyer, 1962).
All the children who failed to pass through program 3, or who com-
pleted it with many errors, ran into trouble at approximately the same
place-somewhere bet,ween brightness levels 15 and 17, when the bright-
ness cue became difficult to use. Eighteen children tried program 3. With
one exception (see below) all the children above 4 years of age climbed
through the fading increments either with no errors or only one error.
Seven children were younger than 4 years. Of these, four reached the
end of the program. The youngest was E.F.C., age 3 years (Fig. ll), and
the next was 3 years and 2 months; both of these children made errors
on only two slides. M.B., age 3 years and 3 months (Fig. l3), and one
other child, age 3 years and 5 months, had similar difficulties but both
did break through. The program was not adequate for the remaining
three children younger than 4 years old.
The one child older than 4 years (4G) who did not get through the
program was severely retarded as a result of a rare metabolic deficiency.
This child’s difficulty turned out to be an inadequate reward. M & M’s
maintained his performance only for short periods of time and then he
stopped working. The breakthrough came when we discovered that an
effective reward for him was the opportunity to feed M & M’s into
the experimenter’s mouth. When he could dispose of his candies that
way, he went through the program quite successfully.
E. The Fourth Teaching Program

The third version of the program was still not as successful with the
younger children as we hoped it would be, but it seemed to be at least
now
I
ONE MINUTE
FIG. 14. The performance of two normal siblings on the fourth circle-ellipse
program. A.P.’s age was 2 years and 11 months; D.P. was 6 years and 8 months.
as effective in 24 slides as program 2 had been in 80 slides (Hively,

1964b). Emboldened by this, we shortened the program still further.
As Fig. 7 indicates, we removed brightness increments 2, 4, 6, 8, and 10,
thereby greatly increasing the rate at which the incorrect keys became
brighter in the early part of the program. The rest of the program
remained the same as the third version, The children now required only
19 slides to get through the program, and again the final 3 slides repre-
sented the criterion performance, with all keys equally bright.
The youngest child who succeeded with program 4 was A.P., age
2 years and 11 months. She made no errors at all, and her record appears
in Fig. 14. The upper record of the figure is her brother, age 6 years and
8 months. Both children gave perfect performances, the major difference
between them being the greater time the younger child required to
make her choices.
Twenty-six children tried program 4. Of these, seven were younger
than 4 years; five of these seven were less than 3 years old and one was
less than 2 years. Only one of them, A.P., went all the way through
the program.
Of the 19 children above 4 years old, three failed to reach the end
of the program. One of these was a child, age 4 years and 2 months, whose
diagnosis was “retardation and juvenile autism.” He had no speech and
was almost unrateable by ordinary psychological tests. H e was assigned.
a mental age of 15 months on the Cattell Infant Intelligence Scale. W e
saw this child five times. In his first session he managed to reach slide 11,
with many errors and with much encouragement and demonstration. In
the second and third sessions we were not able to get him past the first
couple of slides. In the fourth session he reached slide 12 and in the
fifth session he worked considerably more rapidly and reached slide 14.
Again, while far from successful, the program did appear to go a long
way toward making contact with an otherwise unreachable child.
Our groups of children in programs 3 and 4 were certainly not com-
parable, but the data suggest that the more rapid program 4 might have
been somewhat less effective than program 3. Neither, however, gave us
the results we wanted-successful performance by children a t least as
young as 2% years. Our guess was that if we could work successfully with
normal children as young as 2% we would also have considerable success
with older retarded children. Ou r dissatisfaction led us to reevaluate the
procedures we had been using, and this reevaluation finally produced
the program we were seeking.
T h e review went all the way back to C.M., our very first subject,
before we hit on a promising lead. C.M. had been given some preliminary
experiences before he went through program 1. We had completely
underestimated their importance and had neglected them with all sub-
sequent subjects. I n programs 1-4 we tried to accomplish two things
simultaneously. Starting with a circle on a bright key as the correct
choice and seven dark keys as incorrect choices, we gradually increased
the brightness of the incorrect keys; at the same time we gradually faded
the flat ellipse onto the incorrect keys. C.M., however, had first gone
through a program in which we gradually brightened the incorrect keys
without superimposing any form on them. I n essence, C.M.’s task had
been broken u p into two simpler parts, “background fading” and
“ellipse fading.” All the other subjects had been asked to learn both
components of the final discrimination at the same time. Therefore, in
our next revision of the program, we separated these two aspects of the
learning process.
We realized at this time that the concurrent fading of both the back-
ground and the ellipse had an additional consequence that was probably
responsible for many of our failures. Although the ellipse faded in gradu-
ally, it was clearly visible to the children long before its background
brightness equalled that of the circle. Because of this, the children could
continue to use the brightness cue long after the ellipse was clearly
available to them. T h e children needed to look only for the bright key;
they did not even have to observe the circle. Since it was possible for
them to ignore the forms, we can assume many children did. T h e circle
and ellipses were present, but irrelevant. We were, in fact, teaching the
children to ignore them. When the background fading eventually made
it impossible for them to use brightness, many of the children could al-
ready have learned not to pay any attention to the forms. Separating the
background fading and the ellipse fading eliminated this deficiency in
our program.
Similar reasoning may help to explain C.M.’s difficulty with program 1.
Although he had previously learned with ease to select the bright key
with a form on it from equally bright keys without forms, program 1
reestablished a discrimination based on brightness alone. C.M. experi-
enced great difficulty when brightness became difficult to use as a basis
for correct choices. His previous learning may have helped him finally
to attend to the forms again. T h e ease with which C.M. had gone through
the original background-fading program encouraged us to divide the
circle-ellipse discrimination into two stages.
F. The Fifth Teaching Program

As before, the first slide of the new program presented the subject
with a circle on one bright key and seven dark keys. During the first
A C D
FIG. 15. Schematic illustration of a few steps in the background-fading portion of

program 5. T h e correct key always had the circle on a bright background. T h e in-
correct keys were dark at first (A) and gradually became brighter (B, C, D).
eight slides we gradually made the incorrect keys brighter until they
finally equalled the correct. key, but this time no ellipses appeared.
Since no forms appeared on the incorrect keys during this phase we
called it “background fading.” T h e incorrect keys gradually became as
bright as the background of the correct key (Fig. 15). I n order to make
180 Miirray Sidman and Lawrence T . Stoddard
a correct choice, the child had to look for the only key with a form
(“something”) o n it. T h e next step was to teach him to look for a
particular form (circle).
Beginning with slide 9, we gradually introduced the flat ellipse on
the fully bright incorrect keys. T h e ellipses were very faint gray at first,
almost invisible even if one knew to look for them. T h e ellipses became
darker from trial to trial and the contrast between ellipse and background
gradually increased. By slide 17 the only difference between correct and
incorrect keys was the form-circle o r ellipse. We called this phase
“ellipse fading” (Fig. 16).
FIG. 16. Schematic illustration of a few steps in the ellipse-fading portion of pro-
gram 5. The ellipses appeared gradually on the bright backgrounds of the incorrect
keys. (The ellipses were not actually dotted: they were drawn that way here for con-
venience in reproduction.)
Six subjects went through program 5. T h e youngest, 2 years and 3

months old, made little progress, although he did get through the
program after its next revision. Complete data for the other five children
appear i n Fig. 17. T h e encircled numbers indicate those slides on which
the children made “first-time errors,” that is to say, those slides to which
they responded incorrectly the first time they reached them. Errors after
a backup are not numbered, but there were few of these. T h e arrows
at a indicate correct choices on slide 8, the end of background fading. At
this point all keys had become equally bright but only the correct key
had a form (circle) on it. Only one child, M.J.M., made an error in this
part of the program. T h e arrows at b indicate the correct choice on slide
17, the end of ellipse fading. O n this slide the only way the child could
make a correct choice was to distinguish circle from ellipse. Slides 18, 19,
and 20, the final three, were simply repetitions of this criterion per-
formance.
Starting at the bottom of Fig. 17, the youngest child, T.E.S., was 2
years and 9 months old. T h e next two, F.P.R. and F.D.Z., were 4 years
and 7 months and 4 years and 10 months, respectively; the two oldest
children, M.J.M. and S.F.L., were 5 years and 4 months and 5 years and
9 months, respectively.
All the children went through the two-stage fading process with little
difficulty. However, the sample is small. An interesting observation is
the consistency of the errors on slide 13. All but the oldest child made
PKOGKAMMING PERCEPTION AND LEARNING 181
at least one error when they reached slide 13, and he made his error on
slide 12. In no instance were the errors serious-none of the children
“hung up”-but we had obviously created a difficult spot in the ellipse-
fading series. S.F.L., even though he did select the circle on slide 13,
made a revealing comment, “Oh, no! All of ‘em! Wait a minute, there’s
one?” S.F.L.’s comment suggested that he had not observed the ellipses
until slide 13 came up, and at that point they seemed suddenly to appear
on all the keys. It seemed reasonable to suppose that we had made too
large a jump from slide 12 to slide 13 in fading the ellipse on to the
ONE MINUTE
FIG. 17. Thc performance of five normal children on the fifth circle-ellipse program.
The arrows at a indicate the end of background fading; the arrows at b indicate the
end of ellipse fading. The encircled numbers are the slides on which the children made
“first-time” errors. See the text for the childrcn’s ages.
bright background of the incorrect keys. We therefore made a simple

revision, inserting an intermediate fading increment between slides 12
and 13. At the same time, we removed one of the criterion slides from
the end of the program.
G. The Sixth Teaching Program

Program 6, then, was the same as program 5 from slides 1-8, the
Imkground-fading series; ellipse fading now occupied slides 9-18 instead
of 9-17. Slides 19 and 20 were simply repetitions of the criterion perform-
ance. Forty children went through program 6, and like the other pro-
grams, this one was highly successful with children over 4 years old. Of
interest here is the performance of children younger than 4 years.
182 Murray Sidrnnn and Lawrence T . Stoddard
Eight children younger than 4 went through program 6 . Of the eight,

only one child, age 2 years and 4 months, failed to reach the criterion
performance. All the others learned the circle-ellipse discrimination.
Another child, also 2 years and 4 months, made nine first-time errors,
two during background fading and the rest during ellipse fading, but
a recorder failure prevented us from showing his record. T h e performance
of all the other children younger than 4 appears in Fig. 18. T h e youngest
ul
Y
0
0
0
I-
P
a
0
0
FIL. 18. The performance of eight childrcii on the sixth circle-ellipse program. n
indicates the end of background fading, and b indicatcs the end of ellipse fading. The
encircled numbers are the slides on which the childrcn made “first-time” errors. For
tlctailq on the children see the text.
of these, W.J.B., was 2 years and 7 months; N.E.L. was 3 years old;
M.F.R., 3 years and 1 month; P.L.F., 31/, years; J.D., 3 years antl 8 months;
and J.M.N., 3 years antl 9 months. T h e two upper records in Fig. 18 are
those of older children who are especially interesting, antl we sh;ill return
to them below.
Slide 13 was atltled to the program to make the ellipse fading more
gradual a t a critical point; slide 14 in program G is the same ;IS slide 1.3
of Ilrogram 5. T h e addition of the new fading increment did 1101 reall)
clear u p the children’s problem. Most of them made errors either on the
PROGRAMMING PERCEPTION A N D LEARNlNC 183
same slide as before or on the new one. I n n o instance, however, did this
cause mole than a temporary stumbling block to any child’s progress, and
we felt that the difficulties involved in preparing an additional inter-
mediate fading increment would be too great to warrant further revision
at this point. T h e major finding here is that, for the first time, the
program was almost 100% successful in teaching children younger than 4
years of age to discriminate circles from ellipses. Tw o of the children were,
in fact, younger than 3, and N.E.L., who gave the best performance of all,
was exactly 3 years old.
A.J.M., age 4 years and 5 months, was one of a large class of phys-
ically and mentally underdeveloped children whose neurological and
other medical examinations yield largely negative findings. When he
went through conventional psychological tests, A.J.M. received a low-
normal IQ, in the high 80’s. T h e psychologist who tested him reported,
“He appeared not accustomed to working things through himself. H e
could be persuaded to complete tasks but his inclination was to ask for
help immediately when he found anything difficult.” A.J.M.’s rapid and
almost errorless progress through the circle-ellipse program suggests that
w i t h adequate reward for correct performance, and with a teaching
method designed to ease the transitions from one stage of a task to
another, he can avoid the obstacles that slow him down when he is asked
to learn by conventional methods.
D.M.’s record, the topmost one in Fig. 18, also suggests the potential
effectiveness of our new program. D.M.’s medical diagnosis was ‘micro-
cephaly and mental retardation.” H e was 8 years old, but standard psy-
chological evaluation gave him a mental age of 4 years and 8 months and
an IQ in the 50’s. On our program D.M. learned the initial brightness
discrimination without any errors. He did make five errors-a large
number-on his way to learning the circle-ellipse discrimination, but the
backup procedure apparently helped him considerably. Furthermore,
he learned to distinguish circles from ellipses in something less than 3
niinittes. We cannot, of course, predict how far children like D.M. might
eventitally climb; this result suggests that hundreds of hours might not
be too great an investment to apply toward developing teaching programs
similar to oitr circle-ellipse program but more immediately relevant to
the demands of everyday living.
H. The Final Teaching Program

Il’e made only one additional minor change in our program for teaching
children the circle-ellipse discrimination. This change was simply the
removal of one slide from the background-fading portion of the program.
Most children passed through this phase of the program without making
an error. A closer look at the data revealed that no child ever made an
error on slide 7 and very few picked a wrong choice on slides 6 and 8. W e
therefore removed slide 7, leaving only seven steps in which the back-
ground illumination on the incorrect keys changed from dark to fully
bright. We also removed one of the three criterion slides at the end of
the ellipse-fading section of the program. From here on the program con-
sisted of only 18 slides, at the end of which the children were able to
discriminate circles from relatively flat ellipses. This is a far cry from
our second program, which required 80 slides and was not nearly so
successful.
Of the next 30 children who went through the program, only 1 failed
to learn the discrimination. Of these 30, 9 were younger than 3 years. T h e
youngest child, 2 years and 2 months, made many errors, particularly
near the end of the ellipse fading, but he did complete the program. T h e
child who failed to complete the program was 23. Of the others, 2 were
2 years and 7 months, 2 were 2 years and 8 months, 2 were 2 years and
10 months, and 1 was 2 years and 11 months.
T h e child to whom the program failed to teach the discrimination had
so little behavior that she was rated untestable by the clinical psychologist.
This child suffered almost continuous myoclonic seizures-brief but
frequent small muscle spasms and jerks. Another of this younger group of
children failed to go beyond the first few slides when we gave him candies
for his correct responses, but when we changed the reward to small sips
of water he went through the program with little difficulty. This em-
phasizes the close relation between the programming techniques and the
consequences of the desired behavior. An effective reinforcement is
necessary. It is our means for making it worthwhile for the child to follow
the instructions our program gives him. But even with an adequate
reinforcement, the child cannot behave appropriately unless we also give
him adequate instructions. This is the function of the program. Without
an adequate reward, therefore, the child has n o reason to follow our
instructions; without adequate instructions, the reward will make n o
contact with relevant behavior, and the child will learn to do something
else, like pressing all the keys in succession until he finally hits on the
correct one.
I. Recapitulation
All the work we have described so far was devoted to the problem of
teaching nonverbal children a simple visual discrimination. We have
presented our methods in great detail became we believe it important
for others interested in working along these lines not to be misled by the
simplicity of the final product. T h e major part of the labor, however, is
PROGFAMMINC PERCEPTION AND LEARNING 185
confined to the program designer-the teacher. Once developed, the

program is time in the bank. With each child who uses it, time is saved
-both his and the teacher’s.
I n the process of developing the circle-ellipse program we learned a
number of valuable lessons. First, consistent with everything we already
knew about behavior in general, the necessity to reward the child ade-
quately for desirable behavior was brought home to us again and again.
It may even be necessary to deprive a child of food or water in order to
establish an effective reward. T h e progress he makes and the information
we secure compensates for the slight discomfort this may cause him.
Given an adequate reward, we must then attend to the details of the
task we want the child to learn. We must analyze this task into its com-
ponent parts, carefully specify the end point we want the child to reach,
and program the components in an orderly progression from simple to
complex, always building on what we have already taught the child to
do (Holland, 1960; Skinner, 1954). T h e technique used in the circle-
ellipse program began with a simple brightness discrimination, which
appears relatively easy for most children (see also Kellogg & Rice, 1964;
Schlosberg & Solomon, 1943). By careful fading we gradually transferred
control over the child’s behavior from the relative brightness of ‘the
correct and incorrect stimuli to the circles and ellipses. At first we at-
tempted simultaneously to fade out the brightness cue and fade in the
ellipses. When this proved to be too difficult for some children we broke
the task into two components. We first faded out the brightness cue and
then faded in the forms. By separating the child’s perceptual task into
these two components we succeeded in developing a successful program.
We learned next that the teaching program should progress as rapidly
as possible. If the children are given too much “practice” at a given point
they are likely to become bored or satiated. Or if the cue we eventually
want them to use is present as an irrelevant stimulus too early in the
program, they may actually learn to ignore that cue and then be unable
to use it when the alternatives become impossible.
We also learned the utility of giving the child extra help by means of
the backup procedure. When the child makes an error, the program auto-
matically reverses to an easier section, giving the child another chance to
get himself started on the correct course.
Finally, we learned to use the children’s errors as an indication that we
had programmed the task inadequately. By constantly revising the pro-
gram, we were able to eliminate consistent errors that kept the children
from progressing.
All these ingredients are part of the repertoire of any good teacher.
What we have done here is to combine them into one package, to use
186 Murray Sidnian a n d Lazureiice T . Stoddard
automatic equipment that has infinite patience and no lapses of memory

or attention, and to have enough confidence in the methodology to
persist until we arrived at a successful end product.
I n none oE the preceding pages have we demonstrated rigorously that
the program was necessary to teach any OE our children to discriminate
circles from ellipses. This demonstration has been made and will be
published more appropriately in a technical journal (Sidman & Stoddard,
in press). We have shown that the program is an effective device
for nonverbal communication. Without any other form of instruc-
tions we have communicated to our children that they are to choose
circles and reject ellipses. T h e program can accomplish this task of
communication reliably within 5 minutes o r less, with normal chil-
dren as young as 2 to 3 years of age and with older retarded chil-
dren. It does not matter that some of our children could talk or that
some of them already knew the difference between circles and ellip-
ses. Our ultimate interest is in preverbal or otherwise nonverbal
children, with whom some other means of communication must be found
if we are going to be able to evaluate their behavioral potential fairly.
Because most of our standard intelligence tests, even the so-called “non-
verbal” tests, depend heavily on verbal instructions, most children who
cannot talk earn a low I.Q. We often doom these children to a n institu-
tionalized or otherwise impoverished existence simply because they do
not speak our language.
T h e philosophy behind our persistent efforts to develop the circle-
ellipse program can no longer be regarded as a n academic issue. It is a
set of methods that work. We have found empirically that in order to
teach effectively we must first prepare ourselves to learn from our chil-
dren. Their errors are a lesson to us. They make mistakes because our
teaching has been inadequate. T h e proof was the progressive elimination
of errors as we revised our teaching program. With successive revisions
the children made fewer and fewer errors, and more of them reached the
point to which we were trying to get them.
Are these statements too sweeping? After all, we have only taught our
children to tell circles from ellipses, a performance which will hardly play
an important role in their lives. We certainly do not propose our program
as a substitute for I.Q. tests. We believe, however, that the method-
ology is general. It can be applied to areas of much more immediate con-
sequence to the children. We shall expand on this notion below. Mean-
while, let us return to the problem we originally set ourselves. Teaching
the children to discriminate circles from ellipses was only a preparation
for a more precise evaluation of their visual perception. T h e sensory
evaluation of children who cannot or do not talk is a vexing practical
problem to neurologists as well as to teachers who want to know whether
I’ROGRAMMING PERCEPTION AND LEARNING 187
a child is capable of learning some of the things he is ordinarily expected

to learn.
T h e circle-ellipse program serves as a substitute for verbal instructions.
It is our way of telling nonverbal children what to look for. We cannot
say to these children, “Pick out the circle” o r “Press the round one.” But
with our final set of 18 slides we accomplished the same thing without
words. When they completed the program, the children were attending
to the relevant stimulus dimensions just as well as children for whom
verbal instructions are appropriate. We are now ready to determine how
fine a discrimination they can make.
111. THE CIRCLE-ELLIPSE DIFFERENCE THRESHOLD
Teaching did not end with the beginning of our testing procedure. As
we have done all along, we started by asking the child to tell us, non-
verbally of course, something he already knew. We increased the difficulty
of our questions in small increments, and we gave the child as much help
as we could. Again, we used a “fading” technique. T h e child started by
picking a circle and rejecting relatively flat ellipses, with rewards for his
correct choices. From one trial to the next we gradually increased the
vertical dimension of the ellipses, making them look more and more like
circles. If the child made a mistake, picking an ellipse instead of the
circle, the backup procedure returned him to an easier step, giving him a
chance to start again with a discrimination he could make. Figure 19A
shows the progression of ellipse sizes we used in our first program. T h e
numbers beside each ellipse indicate the ratio of its vertical to its
horizontal axis. As these numbers become larger, the ellipses look more
and more like a circle. When a child gets to an ellipse that he can no
longer distinguish from a circle he “hangs up” in the program, oscillating
back and forth between slides on which he can tell the circle and ellipses
apart and slides whose ellipses he cannot tell from the circle. T h e ellipse
beyond which he no longer makes a correct choice is his “threshold.”
Because of the backup procedure, the child can advance only one slide
beyond his threshold.
Figure 20 shows the record of a child who went through our first
threshold program without an error. She was a bright 7 year old and her
performance is not unusual. We present the record mainly to illustrate
the program and indicate what an errorless performance looks like. Al-
together we used 10 sizes of ellipses, each size appearing bn eight consecu-
tive slides before increasing. T h e dots under some of the steps in Fig. 20
indicate the slide on which each pictured ellipse first appeared. This child
made no errors as the ellipses became more like circles, although she did
188 Mirrray Sidman and Lnwrence T . Stoddard
take more time to make her selections when she reached the slides contain-
ing the largest ellipses.
With only a few exceptions, all the children went through the circle-
ellipse program before we tested their threshold. As we indicated in Fig.
20, the original threshold series contained eight presentations of each of
10 ellipse sizes, making a total of 80 slides in that series. Added to the 72
A
1 .oo 1.oo
0
.90 0 .89
X
-
cn 0 .84
-0.86
- 0.83
- - 0.80
4 .80 0 .79
- 0 .77
a 0 .74 -0 .74
0
2 .70 --0 .70
\
E ’ 0 .66 -0 .66
cn ’ 0 .61 .-0 .61
.60
U - 0 .57
a ’ 0 .53 -0 .53
.SO
E .0 - 0 46
E .46
.40 .o.39 -0.39
.30 . 0 3 1
FIG. 19. The progression of ellipse sires in the first (A) and second (B) threshold
series. The number beside each ellipse is the ratio of its vertical to its horizontal axis.
The horizontal axis never changed.
or 80 slides of our original discrimination programs, this constituted a

lengthy procedure. As we became skeptical of the old maxim that “prac-
tice makes perfect,” and reduced the length of the discrimination pro-
gram, we also reduced the length of the threshold series. We cut down
the number of slides per ellipse size from eight to three. I n addition, we
began to increase the ellipse sizes during the course of the fading
program. We had found that children rarely made errors on ellipses
whose axis ratio was .53 or less.
PKOGKAMMING PERCEPTION A N D LEARNING 189
Figure 21 shows the main features of the total series of slides and
demonstrates a phenomenon from which we learned a valuable lesson.
T h e record, made by a normal child, age 5 years and 2 months, will repay
some detailed attention. T h e small arrows indicate the slide on which
each pictured ellipse appeared. At the start of the fading program
(program 3) the correct key was fully illuminated and contained a circle.
T h e .31 ellipse, not yet visible, was gradually faded onto the incorrect
keys. On slide 9, the gradually fading ellipse was changed to a slightly
Fic;. 20. The performance of a normal 7-year-old child on the first threshold pro-
gram. The dots indicate the slide on which each of the pictured ellipses first appeared.
larger one, with an axis ratio of .39. On slide 17, when fading was almost
complete, the ellipse ratio was increased to .46. By slide 22 the fading
process was over, the ellipses were completely faded in, and the ellipse
size increased to .53. This child, J.A.P., made no errors through the
ellipse-and-background-fading program and went through the 3 criterion
slides with no difficulty.
T h e threshold series began on slide 25, with the positive and negative
keys indistinguishable except for the presence of circles or ellipses. T h e
.61 ellipse was introduced at this point. J.A.P. made her first error on
slide 26, at a, returned to slide 25, and then moved rapidly to slide 28.
On slide 28, at b, the ellipse size had jumped to .66, and the child’s
190 Murray Sidman and Lawrence T . Stoddnrd
error on this slide returned her to slide 27, which contained .61 ellipses.
T h e child made three errors on this slide, having made none on her first
exposure to it, returned to slide 26, and then moved successfully to the
.74 ellipse. She made several errors at d , the third slide of the series con-
taining .74 ellipses. At e, she made three errors on the first slide contain-
ing .79 ellipses, returned to the .74 ellipses, and then moved with little
a 00
baa@
c 0000
d@@@@@@
e0000
f 000
so000000
hO@@@@@
i 0000000 s 0000
i 000000 t 000000
FIG.21. T h e performance of a normal child, age 5 years and 2 months, o n the third
circle-ellipse program and on the threshold series. T h e arrows indicate each slide on
which the ellipse size changed, whether because the child advanced or backed up. Note
that the ellipse sizes began to change even during the fading program: fading was
complete on slide 22. T h e key matrix is schematically illustrated at the upper right,
and the keys are numbered 1-9. T h e letters indicate those trials on which the child
made at least one wrong choice, and in the upper section of the figure the numbers
beside each letter show the succession of keys the child pressed. T h e final number on
each line identifies the correct key on that trial.
difficulty to f, her highest point, which was the first slide containing .89
ellipses, At f, J.A.P. had advanced past the last slide that contained .84
ellipses; since she did not advance through the .89-ellipse step, her
circle-ellipse threshold was 34.
Starting with slide 40, at f, the child then made errors on 11 consecutive
slides, backing down all the way to the .G6 ellipse, at p . She was able to
retiirn to the .74 ellipse, at q, but continued to select this ellipse (r, s, t )
and was not able to pass beyond it.
This record is not the first instance we have seen in which “errors
create errors.” T h e trials from f to p , in which J.A.P. backed down
through the program, making errors where she had previously been
successful, are striking but by no means unusual examples. Additional
details from this child’s record illustrate clearly what she was doing. I n
the upper right portio? of Fig. 21 we have schematically illustrated the
matrix of keys, numbered from 1 through 9. In the two columns a t the
left of this illustration we have listed all the trials, from a through t, on
which the child selected an ellipse at least once. Beside each letter we
have indicated the sequence in which the child pressed the keys on that
particular trial. T h e final key in each row is the correct key for that trial.
For example, at a, J.A.P. first pressed key 6, an ellipse, and then pressed
key 9, the circle. Note that keys 6 and 9 are adjacent.
At b, the child pressed key 2 twice in succession before going to key 6,
which was correct. At c, she first selected key 6-which was correct on the
previous slide-and then selected three adjacent keys in succession,
numbers 8, 7, and 4. Th e pattern of moving around the outside of the
matrix, selecting adjacent keys until she came to the correct one, was
already beginning to develop. At d, J.A.P. moved from key 7 to key 4,
skipped key 1, and then continued in the same direction through keys
2, 3, 6, and 9. This pattern of responding became even more striking
after the child reached her high point at f . At g she started on key 1 and
moved along on consecutive keys, making six errors before arriving at
key 3, which contained the circle. T h e child continued with similar pat-
terns all the way down to p , at which point she jumped directly from key
9 to key 3, apparently finding i t easier again to use the difference between
circle and ellipse as her basis for choice. O n trial k, the child made 10
errors before finally selecting the circle. This large number of errors-
more than the actual number of keys-occurred because J.A.P. temporar-
ily abandoned her systematic exploration, jumping from key 8 to key 4,
from key 1 to key 3, and from key 3 to key 9 before she finally returned
to her old system of picking adjacent keys. O n trials I and 0, the child’s
system worked quite well simply because the correct key happened to be
adjacent to the one she had selected first.
I n discussing the original circle-ellipse fading programs, we had to infer
that the children’s errors came because they reached a difficult portion
of the fading sequence. In the threshold series, this inference becomes
more obvious. T h e positive and negative keys were actually becoming
less discriminable as the child progressed, and the series was designed so
that the child would eventually reach an ellipse he could no longer
192 Murray Sidman and Lnwrence T . Stoddard
distinguish from the circle. When the child reached this point he found
that his former criterion for choice no longer worked. It no longer
sufficed to look for the circle or for whatever aspect of the forms he had
been observing. All the stimuli had become “circles.” Reasonably enough,
he changed his criterion. Two substitute criteria that we could easily see
in the children’s data were (a) to select the key which had been correct
on the previous trial, and (b) to develop a fixed pattern of moving around
2
0
I-
2a
a ?.LA
8 6/11/64
& .JV
.31
I
ONE YINUTE
FIG. 22. The performance of a normal child, age 4 years and 5 months, on the
fourth circle-ellipse program and on the threshold series. T h e arrows indicate each
slide on which the ellipse size changed, whether because the child advanced or backed
up. Note that the ellipse sizes began to change even during the fading program; fading
was complete on slide 17. The key matrix is schematically illustrated at the upper
right, and the keys are numbered 1-9. The letters indicate those trials on which the
child made at least one wrong choice, and in the upper section of the figure the
numbers beside each letter show the succession of keys the child pressed. The final
number on each line identifies the correct key on that trial.
the keys until reaching the correct one. Not all children selected as
efficient a pattern as J.A.P., nor was a consistent pattern always evident to
us, but it is reasonable to suppose that some such substitute process was
responsible for the frequent occurrence of errors on slides to which the
children had previously respondecl correctly.
The major point to be made here is that repetitive errors and “false”
hypotheses are not necessary to the learning process. Careful program-
ming can prevent errors from occurring, a t least until the child reaches
the limits of his capability. Once the program produces an error, the child
1’KOC.RAMMING PERCEPTION AND LEARNING 193
often responds by changing his strategy and no longer even observes the
relevant cues until they again become so obvious that he can no longer
ignore them.
Figure 22 is the record of a child, age 4 years and 5 months, whose
performance was strikingly similar to that of J.A.P. This child went
through the threshold series after we had decreased the length of the
original fading program (program 4). Her errors in the part of the
threshold series that was difficult for her caused P.L.A. to abandon the
circle-ellipse discrimination in favor of systematically exploring the keys.
Like J.A.P., her threshold was at the .84 ellipse.
Figure 23 is the record of a child, R.D.C., age 5 years and 3 months,
who had considerably more difficulty than usual with the fading program
.W
.N .74 .m .n .n
z
0
I- .uo l d of c.cwr
Ye .Y $b R.D.C.
L xyu
a
0 1 39
31
ONE MINUTE
FIG.23. The performance of a normal child, age 5 years and 3 months, on the
fourth circle-ellipse program and on the threshold series. The ellipses are identified by
their axis ratios.
(program 4) and who settled down to a relatively stable threshold

performance without backing very far down through the series. We were
not able to detect any systematic pattern of errors for R.D.C. She had no
obvious neurological problem, but it may be of interest that she had been
brought up in a remarkably destructive home situation and was even-
tually relocated in a foster home. Her finest circle-ellipse discrimination,
the .74 ellipse, was considerably below our norms for her age level. It
is possible, however, that this reflects a general behavioral deterioration
rather than a basic perceptual deficit.
I n the records we have just discussed (Figs. 21, 22, and 23) each ellipse
in the threshold series appeared on three consecutive slides. Only when a
child made an error on the first of a set of three slides did the backup
procedure return him to a more discriminable ellipse. For example, in
Fig. 22, P.L.A. made an error at a, the first slide on which she was exposed
to the .74 ellipse. T h e error immediately returned her to the .66 ellipse,
and this little bit of help sufficed to get her past the next step. However,
194 Miirray Sidrnaii a n d Lawrence T . Stoddard
at b her errors came on the third slide of the set that contained .84
ellipses. Her error returned her to the preceding slide, which contained
the same ellipses, and she made four errors on this slide (at c), where she
originally had had no trouble. I n this instance, she could receive no help
from the backup procedure. T h e next change we made, therefore, was to
reduce the number of slides in the threshold series to two per ellipse size.
M'e also made some additional changes at this point: As we described
earlier, (programs 5 and 6), we broke u p the fading program into two
sections, first fading the bright background on the incorrect keys and then
fading in the ellipses. We also eliminated the 3 1 ellipse and started with
the .39 ellipse. I n addition to reducing the number of slides per step in
the threshold program we also inserted some intermediate ellipse sizes
PAS.
1I/Wb4
ONE MINUTE
FIG. 24. The performance of a normal child, age 5 years and 7 months, on the final
circle-ellipse program and on the revised threshold series.
and added some that were closer in shape to the circle. These changes are
illustrated under B in Fig. 19 on page 188. This figure compares the
original and the revised threshold series. We added five elipses at the
upper levels, with axis ratios going from .91 to .985. We also inserted
intermediate ellipse steps at .57 and .70. In addition, from .74 to .89, we
replaced the original four ellipses with a more gradually increasing
sequence of six.
Figure 24 shows the performance of a somewhat brighter than average
child, age 5 years and 7 months, on the revised series. T h e arrows again
indicate each slide on which the ellipse size changed, whether as a result
of the child's progress through the series or as a result of the backup after
he made errors. This child was not able to distinguish the .91 ellipse from
the circle, but his threshold, at 39, was one of the highest achieved by
any child at this age level. P.A.S. was extremely persistent and continued
to oscillate around his threshold level even in the face of frequent errors.
He actually went on for a considerably longer period of time, but we have
not included the rest of his record. Most children ask if they can leave
after making many fewer errors than P A S . did.
This child’s long series of reversals back from the .9l to the .80 ellipses,
a pattern also observed in the records of J.A.P. and P.L.A. (Figs. 21 and
22), illustrates anew the necessity for devising a teaching procedure that
will eliminate errors or postpone them as long as possible. This is espe-
cially true when we are trying to test a child’s perceptual capacity. When
the child makes errors, he abandons the perceptual task we have set for
him and does something else instead. From that point on we cannot be
sure of our test’s validity. T h e changes we made in the threshold series,
as illustrated in the ellipse progression of Fig. 19 and the record of
Fig. 24 (P.A.S.), produced results that strongly support the necessity for
careful programming even in a testing procedure. After we made these
changes our whole set of age norms for the circle-ellipse threshold had
to be revised upward. For example, no child between the ages of 2 and 3
years who went through our original threshold series was able to go
beyond the .79 ellipse. After revising the series, antl making the other
changes described above, 50% of the children in this age range were able
to go beyond the .79 ellipse. I n the 3- to 4-year range, the upper 50oj,
of the children in the first series reached their threshold at the .79 or .84
ellipse; with the revised series, the upper 50% of the children in this age
range attained thresholds ranging from 3 3 to .89. Similar changes OC-
curred at successively higher age levels.
We now have an experiment under way explicitly to investigate the
effects of the rate of progression through the ellipse series. There is
already considerable evidence in the literature that a more gradual pro-
gression actually produces finer difference thresholds. William James
(1890) noted this phenomenon as long ago as 1890; Pavlov (1927) demon-
strated it experimentally with dogs as subjects and circles and ellipses as
stimuli; similar findings have been reported with human subjects and
for other types of stimuli (Baker & Osgood, 1954; Lashley, 1938; Lawrence,
1952). We expect simply to confirm these findings under our particular
set of conditions. T h e lesson has been available to us for a long time antl
its neglect has undoubtedly been responsible for the underevaluation of
many children’s sensory capacities as well as the abandonment of many
children on the grounds that they could not be evaluated at all.
We present one more record in Fig. 25, to illustrate the present state of
our threshold series, with a subject who was able to progress almost
through the whole series. Note that we now use only one slide for each of
the first few ellipses in the threshold series. W.G.L. was a normal 16 year
old who received pennies for his correct choices. H e was able to get
through the .95 ellipse successfully but could not discriminate the .97
ellipse from the circle. This young man did not need the brightness and
ellipse fading programs-we could simply have told him to pick out the
circle-but our program was designed as a means of nonverbal communi-
cation.
We have referred briefly to age norms for the circle-ellipse discrimina-
tion. In their present incomplete form, the normative data indicate a
rapid increase in discriminative ability from ages 2 to 6, and then only a
slight increase above age 6. Does this mean that the circle-ellipse dis-
crimination reflects a developmental process that is not complete until a
child reaches 6 years of age? While many neurologists suggest that the
End of
W.G.L.
ONE MINUTE
FIG.25. The performance of a normal 16 year old on the final circle-ellipse program
and on the final threshold series.
nervous system continues to develop even beyond age 3, there is n o

conclusive neuroanatomical evidence in favor of this, particularly in the
visual system.
If a still-developing nervous system is not responsible for the finer circle-
ellipse discrimination with increasing age, where else may we look for
an explanation? Perhaps an answer is to be found by looking more
closely at the behavior involved in distinguishing these two stimuli. O n
what aspects of the two forms do the children actually base their discrimi-
nation? We have been calling the forms circles and ellipses, as if this
were a sufficient specification. Perhaps, however, the children are looking
at things other than the overall shapes. For example, the vertical axis, or
height, of the circle is greater than that of the ellipse; the two stimuli also
differ in area; the sides of the ellipse are sharper than the circles; the
ellipses are flatter at the top and bottom. Children, and adults, may use
any or all of these cues in making their discriminations. Perhaps the
younger children fail to make finer discriminations simply because they
have not yet learned to attend to such details of their environment as
heights, areas, and degrees of curvature. Suppose we explicitly taught

them these fine discriminations, devising programs not to teach the gross
circle-ellipse discrimination, but rather to teach such discriminations as
height, area, and curvature? Would this then increase their circle-ellipse
thresholds?
At present we cannot answer this interesting question, but it is impor-
tant enough to form a major part of our future plans. T h e answers will
not only clarify some basic relations between behavior and the develop-
ing nervous system, but will have practical results as well. For example, if
a child performs at a lower than normal level in the circle-ellipse dis-
crimination, does it mean he has poor vision, or has he simply not yet
learned to attend to the fine details of his environment? Most of the
institutionalized retarded children we have tested have scored well below
the norms for their ages. Do they need eyeglasses or do they need more
training in perception?
We have described the development of a nonverbal technique for teach-
ing children to discriminate circles from ellipses, and a combined teaching
and testing procedure, still nonverbal, for determining the quantitative
limits of their ability to distinguish these forms. It is now possible for
us to secure information about visual function in children who are
otherwise unreachable. While our methods have been applied only to a
single aspect of visual perception, there is no reason why they could not
be extended also to visual functions such as acuity, brightness thresholds,
flicker fusion, etc. T h e techniques are potentially valuable not only for
testing sensory capacities, but for teaching the children skills that are
more relevant to their everyday necessities. I t should be possible to use
letters, numbers, words, colors, sounds, etc., and thereby lay the founda-
tion for programming more advanced materials (Holland & Matthews,
1963).
IV. THE REVERSAL OF THE DISCRIMINATION
To demonstrate the generality of our techniques we set out to develop

a program to teach the children, without verbal instructions, to reverse
the discrimination they had already learned. We had taught them LO
select circles and reject ellipses; our task now was to teach the same
children to select ellipses and reject circles. If we could teach nonverbal
children this discrimination reversal, and if they could learn the reversal
without even making an error in the process, this would indicate the
feasibility of wider application of our methods.
Our plan involved two major stages. We would gradually transform the
positive stimulus from a circle to an ellipse, and, conversely, we would
198 Murray Sidtnnn a n d Lnwrence T. Stoddard
gradually transform the negative stimuli from ellipses into circles. But
if we did both of these simultaneously, there would be a stage in the
program where the positive and negative stimuli would be exactly the
same, and the chilclren would have no basis for a correct choice. Ther e
had to be an intermediate step i n which either the positive or negative
stimuli would temporarily assume some third shape, thereby assuring that
the positive and negative stimuli were always different. To determine
what this third shape should be, we performed a little experiment. T h e
data of this experiment are also relevant to the problem we proposed
earlier; namely, to what aspects of the circle antl ellipse are the children
a ttentling?
A. The Probe Series

O u r initial problem was to determine which stimulus to transform
first, the positive circle or the negative ellipse. T h e second problem
was to select an intermediate form to use in this transformation. If we
changed the circle into something else, the new stimulus should be one to
which the children would transfer their choice with little difficulty. If we
were to change the ellipses, we should use a stimulus the children would
reject easily. This is what Kluver (1933) called the problem of “equivalent
stimuli.” What forms are equivalent to-substitutable for-the circle and
the ellipse?
We devised a series of slides almost exactly like the ones to which the
children were already accustomed. We projected a circle on one key and
ellipses on the other seven keys, and the circle was the correct choice.
Every fourth slide in the series, however, was a probe. On these probe
slides we did one of three things: we substituted a new form for the
circle and left the ellipses unchanged; we substituted a new form for the
ellipses antl left the circle unchanged; or we substituted new forms for
both the circle and the ellipses.
None of the stimuli on the probe slides could be called correct or
incorrect. Therefore, the child always advanced to the next slide no
matter which key he pressed on a probe slide. We gave n o reward for the
child’s response to the probe stimuli, nor did the chimes ring. T o prevent
the lack of reward from giving him false clues we instituted a procedure
of intermittent reward during both the probe series antl the preceding
threshold series. T h e child received rewards for only 5OC& of his correct
choices, and the trials on which he received rewards were schetlulcrl in an
irregular sequence.
T h e stimuli we used on each probe slide may be seen at the top of
Fig. 26. T h e number 1 indicates the stimulus that appeared on only one
of the keys, antl the number 7 indicates the stimulus that appeared on
PROGRAMMING PERCEPTION A N D LEAKNING 199
each of the other seven keys. T h e first four pairs of stimuli show the
seven ellipses unchanged, but substituting for the tiicle were a square, a
smaller circle, a triangle, or an X. I n the next three pairs we left the
circle unchanged, but for the usual ellipses we substituted horiLonta1 but
elongated ellipses, vertical ellipses, or squares. I n the final three pairs of
stimuli, we substituted, first, a square for the circle along with triangles
for the ellipses; then, a triangle for the circle and squares for the ellipses;
antl finally we simply reversed the original stimuli, substituting the ellipse
FIG. 26. Responses to the probe slides. Each set of initials represents one child, with
his age indicated to the nearest full year. Each pair of probe stimuli is depicted at the
top. T h e number above each form indicates the number of keys on which it appeared.
T h e first four pairs of stimuli involved a change in the circle, with the ellipses re-
maining the same; the next three pairs invohed a change in the ellipscs. with the
circle remaining the same; the final three pairs involved changes in both the circle
and the ellipscs. T h e checkmarks show which stimulus each child selected on each
probe slide. Double checks indicate that the child was exposed to the probe series
twice. Question marks indicate data lost through apparatus failures. T h e upper num-
ber of each pair at the bottom of the figure shows the proportion of all the children
who selected the form appearing on only one key-the odd form. T h e lower numher
of each pair gives the same proportion, h u t only for those children listed below the
broken line-all younger than 7, except C.M.
for the circle and circles for the seven ellipses. T h e probe slides were
separated from each other by three of the usual circle-ellipse slides, on
which the circle was the correct choice.
T h e probe slides were presented to each child in one of two sequences,
neither of which is indicated by the arrangement of Fig. 26. T h e ordinal
number of each pair of probe stimuli, reading from left to right in Fig. 26,
was: 1, 3, 6, 8, 2, 5 , 7, 4, 9, and 10. T h e second sequence was exactly the
reverse of the first. T h e main feature of these sequences is that in one, the
square antl ellipse (column 1) were presented first in the series antl the
200 Murray Sidinan and Lawrence T . Stoddard
reversed ellipse-circle combination (column 10) was presented last. In the

second series, the stimuli in column 10 appeared first and those in
column 1 last.
Each set of initials represents one child; his age, to the nearest year,
appears beside his initials. T h e stars indicate three children with posi-
tively established neurological difficulties. T h e checkmarks are located
beneath the stimulus the child selected on each probe slide. Double
checkmarks indicate that the child was given both sequences of probe
slides. T h e question marks indicate a few instances in which we lost the
data because of apparatus failures. Beneath each column of checkmarks
are two numbers. T h e upper number indicates the proportion of all
trials on which the children selected the stimulus which appeared on only
one key. T h e lower number indicates the same proportion, but only for
those children whose data appear below the broken line, those younger
than 7. C.M. was included in this group because he was so severely
retarded that he was more appropriately classified with the youngest
children.
Several features of the data in Fig. 26 are worth noting. First, we see
that when a large square was substituted for the circle, keeping the
ellipses unchanged on the other seven keys, all children selected the
square. This was true whether the children saw this probe slide first or
last in the series. In other words, when ellipses remain on the seven
incorrect keys, a large square is equivalent to the circle. We shall make
use of this finding later in our reversal program.
T h e second finding is that it seemed to make little difference whether
we substituted the new stimuli for the circle (first four columns) or for
ellipses (columns 5 through 7). Of special relevance is the probe slide in
which we substituted large squares for the ellipses (column 7). T h e
children younger than 7 were almost as likely to pick the square as the
circle. T h e fact that there were seven squares and only one circle on the
slide undoubtedly contributed to this tendency, but it is also consistent
with the previously noted equivalence of the circle and square.
T h e third notable feature is that almost all the children older than 6
made their choices on the basis of oddity rather than the specific forms-
they selected the one stimulus that was different from all the others. At
ages 6 and below, this tendency was not nearly so marked. Of the eight
children older than 6, all but three used the oddity principle throughout
the series of probe slides. Some might have learned to do this simply be-
cause of their exposure to the probe slides, but this was not always true.
T.P.,age 10, and D.P., age 7, both selected the single ellipse over the seven
circles (column 10) even though they were exposed to this pair of probe
stimuli first. On the other hand, D.P., age 8, who also saw this slide first,
PROGRAMMING PERCEPTION AND LEARNING 20 1
selected the old positive stimulus, the circle. From then on, however, he
followed the oddity principle. S.D.E., the only one of the older group
who did not show any systematic method of choice, was a child who
suffered almost continual minor seizures.
Of the children younger than 7 only N.S., whose data are incomplete,
appeared to operate exclusively according to the oddity principle. T h e
fact that N.S. selected the ellipse on the reversal slide (column lo), even
though she was exposed to this slide first in the series of probes, argues
strongly that she was using the more advanced form of behavior rep-
resented by the oddity principle. D.B. and, to a slightly lesser extent,
R.P.L., appeared to be using oddity as a basis for choice; yet both of
these children selected one of the circles on the reversal slide even though
they saw this slide at the end of the probe series. They perhaps used both
criteria, oddity and shape, selecting a circle whenever it appeared, but
selecting the odd stimulus when there was no circle. Of the younger chil-
dren, only two selected the ellipse on the reversal slide of column 10, as
contrasted with six of the older children.
T h e number of childen in our sample is small, but we may draw the
tentative conclusion that children younger than 7 are more likely to
base their selection on a specific form rather than on the more general
oddity principle. This finding will become relevant when we look at the
reversal program.
A fourth general conclusion we may draw from the data of Fig. 26 is
that children may use any of several aspects of the stimuli in discrimi-
nating the ellipse and circle. For example, their unanimous selection of
the large square over the relatively flat ellipse suggests that the total area
or the height of the stimuli is a controlling factor in their choices. T.E.S.
selected the large square every time it appeared, and chose the ellipse
when it was rotated so that it looked longer than the circle (column 6).
E.F.C. selected the large square almost every time, selected the elongated
ellipse twice, but selected the small circle twice when it was paired with
the usual ellipse. A.P.’s data suggest that she was somewhat confused by
the series of probe slides; she made six errors on the intervening circle-
ellipse slides even though she had previously learned this discrimination
well. Her selection of the triangle over the square in column 8, and the
square over the triangle in column 9, suggests that she was tending to
respond somewhat randomly, thereby selecting one of the stimuli that
appeared on seven of the eight keys. C.M. showed a similar tendency.
T h e data we obtained from the probe experiment permitted us to
move on to our next goal, teaching the children to reverse their original
circle-ellipse discrimination. T h e relevance of the probe data will become
more apparent below.
8. The Reversal Program

Like o u r programs for teaching the children to distinguish circles from
ellipses and for testing their discriminative capacity, the program for
reversing the circle-ellipse discrimination went through several revisions
before it finally carried its message effectively. W e shall not recapitulate
--
how the children’s errors taught us to create a n effective instructional
slide+ 1 7
-
kler
--
#s:
1 716;
- 1 7 1 7
1 0 0 9 DO 23 D O 34 00
2 0 0 10 DO 24 00 35 00
3 0 0 11 D o 25 00 36 00
4 0 a 12 00 26 00 37 00
5 0 0 I3 00 27 00 38 00
6 0 0 14 00 28 00 39 00
7 00 15 00 29 00 40 co
8 1 7 0 16 00 30 00 41 co
17 no 31 00 -
4i 30
18 00 32 01
7 -
43 0 c.
19 00 33 00
20 00
21 CIO
-oc
22
- -
FIG. 27. Schematic representation of the reversal program. The stimulus pairs are
depicted beside the number of the slide on which they appeared. The numbers 1 and
7 show whether the stimuli appeared on one key or on seven kcys. With the exception
of probe slides 5 and 43, the odd stimulus was always positive, and the stimulus on the
other sewn keys was always negative.
device, for the story would be too much like the one we have already
told. Figure 27 illustrates the program at which we finally arrived.
With the exception of slides 5 and 43, which we shall explain below,
the stimuli in the columns headed by the number 1 are the positive stim-
uli and those in the columns headed 7 are the negative stimuli. T h e
numbers, 1 and 7, refer to the fact that the positive stimulus always
appeared on only one key on a given trial, while the negative stimuli
always appeared on seven keys. T h e numbers beside each pair of stimuli
denote the consecutive slides of the program.
I n the first four slides the positive choices are circles, and the negative
choices are flat ellipses. Since most of the children had been tested for
their circle-ellipse thresholds, the first four slides served simply to re-
establish the old discrimination.
Slide 5 is a probe. It presents the task that the reversal program was
designed to teach; one key contains the ellipse and the other seven keys
contain circles. If the child had already learned to make his selection
on the basis of oddity, he would select the ellipse on slide 5, indicating
that he did not require a special program to teach him to reverse the
discrimination. Consistent with the data we had obtained from the
earlier probe series, only a few of the older normal children selected the
ellipse at this point. Since slide 5 was a probe, and not part of the
teaching program, neither stimulus could be called correct or incorrect.
Therefore, the child always advanced to slide 6 n o matter which key he
pressed on slide 5. We gave no reward for the child’s response to slide 5,
and we used the same procedure of intermittent reward for correct choices
during the reversal program that we used with the probe series.
Slide 6 marks the beginning of the reversal program itself. O n this
slide, we began to square the circle, a process that was complete by slide
8. T h e probe data had indicated the equivalence of the square and
circle, and in our first reversal program we had made an abrupt transition
from circle to square. T h e substitution caused no problem for more than
95% of the children, but because a few had trouble at this point we made
the change slightly more gradual by inserting slides 6 and 7.
O n slide 8, then, the children selected the square and continued to
reject ellipses. We were then ready to accomplish the first major part
of the discrimination reversal, substituting the circle for the ellipse as
the stimulus to be rejected. T h e process by which we accomplished
this is illustrated in the second column of Fig. 27, which shows how,
in slides 9-22, we gradually transformed the ellipse into a circle. I n
slide 9 the children selected the square and rejected the ellipses; by the
time they reached slide 22, selecting the square all the way, they had
learned to reject the same circle that had previously been their choice.
Half of the job was done by slide 22. T h e next task was to transform
the square into an ellipse so that the children would end by selecting
the ellipse while continuing to reject circles. T h e third column of
Fig. 27 illustrates the first part of this process. T h e height of the square
was gradually reduced until, on slide 33, it had become a narrow rec-
tangle. T h e reversal was almost complete. Beginning with slide 34, in
the last column of Fig. 27, the corners of the rectangle were gradually
rounded. By slide 40 the rectangle had been transformed completely
into an ellipse, and the children were now selecting the ellipse while
204 Murray Sidman and Lrrwrence T . Stoddard
rejecting circles. Slides 40, 41, and 42 represent the terminal performance
toward which the program was directed.
Slide 43, at the end, is another probe. I t was designed to help clarify
what the children were doing at the end of the reversal program. We con-
sidered two possibilities: they might have learned to select the odd
stimulus, i.e., the one key that was different from all the others; or they
might have learned to select a particular form, the ellipse. Slide 43, like
slides 1-4, had a circle on one key and the flat ellipse on the other seven
keys. If, by the end of the reversal program, the children had learned
to select the ellipse regardless of how many keys it appeared on, then
they would also select one of the seven ellipses on slide 43. If, on the
other hand, the children had learned the more general oddity principle,
then they would shift from the ellipse on slide 42 to the circle on slide 43,
picking the only stimulus that was different. Again, consistent with the
data from the probe series, only a few of the older normal children
selected the circle on slide 43. None of the younger normal children
or the older retarded children who went through the reversal program
indicated that they had learned the oddity principle. All of them demon-
strated that they were responding to a particular form, the ellipse.
Early versions of the reversal program were shorter than the one
shown in Fig. 27. We have already noted the addition of two interme-
diate steps in the process of squaring the circle (slides 6-8). We also
found it necessary to slow down the rate at which the negative ellipse
changed into a circle (slides 9-22). This process occurred much more
rapidly in our early programs, with larger jumps in the ellipse size
from trial to trial, and, correspondingly, with more errors by the children.
Additional research will clarify this process, but it seems likely that the
children’s difficulties in this part of the program are intimately tied
u p with their circle-ellipse thresholds. At some point in the series, depend-
ing on ‘the child’s age, the ellipse becomes indistinguishable from a
circle. If the child’s threshold is reached after a sudden large increase
in the size of the ellipse, he is likely at that point to recognize his old
friend, the circle, and to select it in preference to the square. By chang-
ing the ellipse size more gradually throughout the series we prevented
the sudden emergence of the circle, and with this more gradual trans-
formation, more children continued to select the square and reject the
ellipses.
T h e series in the third column of Fig. 27, in which the square was
gradually transformed into a thin rectangle, also had to be lengthened
in successive revisions of the program. We found that the height of the
rectangle had to be at least as small as the minor axis of the ellipse
before we could start the transformation from rectangle to ellipse. And
PROGKAMMING PERCEPTION AND LEARNING 205
that final transformation also had to be slowed clown. Many more

children made errors when we used only three slides to complete the
transformation from rectangle to ellipse.
We still have many questions to answer about the reversal program.
There is no reason to believe that this is the only, or even the most
efficient way to teach the children to reverse the circle-ellipse discrimi-
nation. We also have yet to evaluate the role played by the children’s
errors in our program as compared with the errors they make in learning
the discrimination reversal without help from any teaching program. We
have, however, demonstrated that our program can succeed as a com-
munication device, that children can learn to reverse their original
criteria for choice without any other instructions.
I n Fig. 28, the bottom record, A, is that of the youngest child, age
2 years and 7 months, who went through the reversal program without
error. D.L.C., who had originally learned to pick circles and reject
ellipses, ended by picking the ellipse and rejecting circles. On slide 5 she
selected a circle, but she went through the program before we had added
slide 43, so we do not know whether she learned the oddity principle.
Record B in Fig. 28 shows another errorless progression through the
reversal program. This record is of particular interest because R.W.F.
was a 16-year-old mongoloid child who, when he was 13 years and
7 months old, earned a mental age of 3% years and a Stanford-Binet
IQ in the low 30’s. R.W.F., who completed the program relatively
rapidly, selected the circle on slide 5 and the ellipse on slide 43, indi-
cating that he did not learn the oddity principle.
Record C in Fig. 28 is that of a child, B.D., who made several errors
on his way through the reversal program. B.D. was a 7G-year-old child
who had earned a Stanford-Binet mental age of 3% years and an IQ
in the low 40’s. Both the neurologist and the psychologist who examined
him recommended special schooling and institutionalization. B.D. had
his major troubles with slides 8 and 14 of the program and made one
additional error on slide 23. H e was one of the few children who selected
the ellipse over the square on slide 8. It is interesting to note that the
ellipses on slide 14, where B.D. had considerable difficulty, had an axis
ratio of .74, and that his circle-ellipse threshold was at .77. On the probe
slide, B.D. also indicated that he had not learned the oddity principle.
I t is possible that B.D. might have learned to reverse the discrimination
just as easily if we had simply presented him with the reversal problem
without the teaching program. This general question is currently under
investigation in our laboratories with severely retarded institutionalized
children as subjects.
Although the data reported here do not demonstrate that our program
is necessary to teach nonverbal children to reverse the circle-ellipse

discrimination, they do indicate the utility of the technique. If a non-
verbal child can be taught to reverse this discrimination, there is no
obvious reason why a similar method could not be used to transfer the
discrimination t o other materials as well. Starting, let us say, with num-
bers, we might teach a child to select the number 1 and reject each of
the other digits; we might then transform the 1 into, let us say, a 7, and
go through the process again. We could continue until the child had
FIG.28. The performance of three children on the reversal program. For details
regarding D.L.G., R.W.F., and B.D., see the text. Errors were made by B.D. on slides
8, 14, and 23. The other numbers indicate the end points of subsections of the
program (see Fig. 27).
learned to distinguish all of the numbers from each other. This dis-
crimination is a prerequisite before the child can go on to learn the
names and the uses of numbers. Other stimuli, like letters, colors, geo-
metric forms, etc., could be taught i n the same manner. A major signifi-
cance of our work with the circle-ellipse discrimination lies in the promise
of the method as a generally effective teaching device for many kinds of
material and for children with whom we are unable to make contact
through conventional methods. C.M., our microcephalic subject who was
the first to try the original circle-ellipse program, also succeeded in
reversing the discrimination. It remains to be seen how far we can
carry him and others like him.
PROGRAMMING PERCEPTION .4ND LEARNING 207
ACKNOWLEDGMENT
T h e authors are indellted to several colleagues and assistants for their advice and
help: Ronald Ray, Barbara Ray, George Toomey, Mary Carpenter, Julia Mount, Judith
Klinger, Christine Palmer, Mary Evelyn Porter, Dudley Templeton, and Joyce Medverd.
REFERENCES
Baker, R. A., & Osgood, S. W. Discrimination transfer along a pitch continuum. J.
e x p . Psychol., 1954, 48, 241246.
Bijou, S. W. Systematic instruction in the attainment of right-left form concepts in
young and retarded children. I n J. G. Holland and B. F. Skinner (Eds.), A n
analysis of the helrariioral processes involrred in self-imtruction w i t h teaching
machines. Harvard Univer. Teaching Machine Project, Final Report, 1964.
Ferster, C. B. Positive reinforcement and behavioral deficits in autistic children. Child
Develpm., 1961, 32, 437-456.
Ferster, C. B., & DeMcyer, M. T h e development of performances in autistic children in
an airtomatically conlrolled environment. J. chron. Dis., 1961, 13, 312-345.
Ferster, C. B., & DeMeyer, Xi. A method for the experimental analysis of the behavior of
autistic children. Amer. J. Orflropsychiat., 1962, 32, 89-98.
Hively, W. Programming stimuli in matching to sample. J . exp. anal. Beliav., 1962,
5 , 2i9-298.
Hively, W. A multiple-choice visual discrimination apparatus. J. e x p . anal. Behav.,
1964, 7 , 387-389. (a)
Hively. W. I’aramctric experiments with a matching-to-sample program. I n J. G .
Holland and H. F. Skiliner (Eds.), A n analysis of the behavioral processes iniiolued
in self-instruction with teaching machines. Harvard Univer. Teaching Machine
Project, Final Report, 1964. (b)
Holland, A. L., & Matthews, J. Application of teaching machine concepts to speech
pathology and audiology. Asha, 1963, 5 , 454-482.
Holland, J. G. Teaching machines: an application of principles from the laboratory.
J. exp. anal. Behnv., 1960, 3, 275-287.
Holland. J. G. Evaluating teaching machines and programs. Teach. Coll. Rec., 1961,
63, 56-65. (a)
Holland, J. G. New directions in teaching-machine research. In J. E. Coulson (Ed.),
Prograiirtned learning and contpi~ter-basedinstruction. New York: Wiley, 1961. (b)
James, I$’. Principles of psyclioloa. New York: Holt, 1890. 2 vols. Pp. 505-515.
Kanner, L. Autistic disturbances of affective contact. N e w . Child, 1943. 2, 215-250.
Kanner, L. Early infantile autism. J. Prdiat., 1944, 25, 211.
Kellogg, \V. N., & Rice, C. E. Visual problem-solving in a bottlenose dolphin. Science,
1964, 143, 1052-1055.
Kliivcr. H. Belrarrior mechnnisins in ?,ionkeys. Chicago: Univer. of Chicago Press, 1933.
Krasner. I.., & Ullmann, I.. P. Research in behavior modificntion. New York: Holt.
1965.
I.ashlcy, K. S. T h e mechaitism of vision: XI’. Preliminary studies of the rat‘s capacity
for dctail vision. J. gen. Psycho/., 1938, 18, 123-293.
I.arvrcnce, D. H. T h e transfer of a discrimination along a continuum. J . coinp. physiol.
Psylrol., 1952, 45, 511-516.
Lintlslcy. 0. R. Operant conditioning methods applied to research in chronic schizo-
phrenics. P.ychiat. T-es. Rep., 1956, 5 , 118-139.
Moore, R., 8: Goldiamond, 1. Errorless establishmcnt of visual discrimination using
fading procedures. J. exl’. anal. Beliav., 1964, 7 . 269-272.
Pavlov, I. P. Conditioned reflexes. (Translated by G . V. Anrep.) London and New

York: Oxford Univer. Press, 1927.
Schlosberg, H., & Solomon, R. L. Latency of response in a choice discrimination. 1.
exp. Psychol., 1943, 33, 22-39.
Siciman, M., & Stotldard, L. T. T h e effectiveness of “fading” in programming a simul-
taneous form discrimination for retarded children. J . e x p . anal. Behav. (in press).
Skinner, B. F. T h e science of learning and the art of teaching. Haruard educ. Rev.,
1954, 24, 86-97.
Terrace, H. S. Discrimination learning with and without “errors.” J . exp. anal. Behav.,
1963, 6, 1-27. (a)
Terrace, H. S. Errorless transfer of a discrimination across two continua. J . exp. anal.
Behav., 1963, 6 , 223-232. (b)
Ullmann, L. P., & Krasner, L. Case studies in behavior modification. New York: Holt,
1965.

Sidman & Stoddard 1966 Programming Perception and Learning - Circle Elipse

Uploaded by

Copyright:

Available Formats

Sidman & Stoddard 1966 Programming Perception and Learning - Circle Elipse

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sidman & Stoddard 1966 Programming Perception and Learning - Circle Elipse

Uploaded by

Copyright:

Available Formats

Programming Perception and Learning

for Retarded Children'

I. General Methodology ................................ 152

IV. T h e Reversal of the Discrimination .................. 197

This report describes the beginning of a research program that has

relatively unusual mixture of basic and applied research (Lindsley, 1956).

it a practice to help our children adapt to the laboratory situation. To

FIG. 1. The nine-key matrix.

simply to convince him verbally) that we are to be trusted. Not until

techniques similar to ours yield consistent results in widely differing

FIG.2. A child seated in front of the nine-key matrix.

and mounting them in slides for projection. They need no complicated

11. THE CIRCLE-ELLIPSE DISCRIMINATION

How do you go about getting a nonverbal, severely retarded child to

We started, then, by building on behavior we had already taught

T h e horizontal distance between steps shows how much time elapsed

3 In the photographic piocess, Bourges overlays of varying degrees of opacity were

T o minimize any disruptive effects of the interruption, the first few

represent to us the actual progress of the child, moving upward step by

practice of drawing learning curves loses its meaning when we revise

correct key on one out of eight trials, or 121/2y0. If he eliminated a

C. The Second Teaching Program

came at b, the first slide of level 17. T h e subsequent return to level 16

where in the room, occasionally returning to the keys. The frequent

A more extended report on R.L. will be prepared later, in conjunction

E. The Fourth Teaching Program

as effective in 24 slides as program 2 had been in 80 slides (Hively,

F. The Fifth Teaching Program

FIG. 15. Schematic illustration of a few steps in the background-fading portion of

Six subjects went through program 5. T h e youngest, 2 years and 3

bright background of the incorrect keys. We therefore made a simple

G. The Sixth Teaching Program

Eight children younger than 4 went through program 6 . Of the eight,

H. The Final Teaching Program

confined to the program designer-the teacher. Once developed, the

automatic equipment that has infinite patience and no lapses of memory

a child is capable of learning some of the things he is ordinarily expected

111. THE CIRCLE-ELLIPSE DIFFERENCE THRESHOLD

.40 .o.39 -0.39

or 80 slides of our original discrimination programs, this constituted a

(program 4) and who settled down to a relatively stable threshold

nervous system continues to develop even beyond age 3, there is n o

heights, areas, and degrees of curvature. Suppose we explicitly taught

IV. THE REVERSAL OF THE DISCRIMINATION

To demonstrate the generality of our techniques we set out to develop

A. The Probe Series

reversed ellipse-circle combination (column 10) was presented last. In the

8. The Reversal Program

that final transformation also had to be slowed clown. Many more

is necessary to teach nonverbal children to reverse the circle-ellipse

Pavlov, I. P. Conditioned reflexes. (Translated by G . V. Anrep.) London and New

You might also like