See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/18685897

A behavior modification approach to the

treatment of autistic children

Article  in  Journal of Autism and Developmental Disorders · April 1974

Source: PubMed

CITATIONS READS
57 2,961

3 authors, including:

Robert Koegel
Stanford University
218 PUBLICATIONS   13,703 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

koegel positive behavioral support research program View project

Koegel autism research program View project

All content following this page was uploaded by Robert Koegel on 13 July 2015.

The user has requested enhancement of the downloaded file.

 Journal o f Autism and Childhood Schizophrenia VoL 4, No. 2, 1974

A Behavior Modification Approach to the Treatment

o f Autistic Children 1,2
O. I v a r L o v a a s s
University o f California, Los Angeles

Laura Schreibman
Claremont Men's College

Robert L. Koegel
University o f California, Santa Barbara

h7 the very short span o f 12 years, behavior modification has contributed in a

m a j o r way to the education o f the autistic child, lts ma/or contribution lies in its
d e m o n s t r a t e d effectiveness. A s such, it is the only intervention which has beers
empirically d e m o n s t r a t e d to o f f e r help f o r autistic children. Each child who
u n d e r w e n t treamzent made nzeasurable progress, even though the progress was
slow a n d incremental and f e w children became "normal. "

W h e n w e t a l k a b o u t autistic children w e are d e s c r i b i n g c h i l d r e n w h o m a n i f e s t

s e v e r a l c h a r a c t e r i s t i c p a t h o l o g i c a l b e h a v i o r s (e.g., social w i t h d r a w a l , s e l f - s t i m u l a -
tion, ritualistic behavior, echolalic and psychotic speech, apparent sensory
d e f i c i t , a f f e c t i v e i m p o v e r i s h m e n t ) . W h e n u s i n g s u c h d i a g n o s t i c label o n e is t y p i -
cally c o n c e p t u a l i z i n g autism as a d i s t i n c t e n t i t y ; an u n d e r l y i n g p r o c e s s w h i c h is
s e e n as t h e c a u s e o f t h e s e d e v i a n t b e h a v i o r s . I n d e e d , a g r e a t deal o f r e s e a r c h h a s
f o c u s e d o n t h i s o n e u n d e r l y i n g p r o c e s s as t h e b a s i s for t h e p s y c h o p a t h o l o g y .

~Presented at the First International Kanner Colloquium in Chapel Hill, North Carolina,
October 3 t - N o v e m b e r 2, 1973 and to appear in Child Development, Deviations, and
Treatment. Plenum Publishing Corporation (in press). Portions of this paper have previ-
ously appeared in Chapter VII o f the Seventy-second Yearbook of the National Society for
the Study of Education, Behavior Modification h~ Education, 1973.
2The research in our laboratory was sponsored by USPHS Research Grant No. 11440. from
the National Institute of Mental Health, and EHA Title VI-B, No. 42-00000-0000832/025,
from the California Department of Education to the Office o f the County Superintendent
o f Schools, Santa Barbara, California.
Requests for reprints should be sent to Dr. O. Ivax Lovaas, Department of Psychology,
University of California, Los Angeles, California 90024.
111
Q 1 9 7 4 P l e n u m PuDlisrling C o r p o r a t i o n , 2 2 7 West 1 7 t h S t r e e t , N e w Y O r k , N . Y . 1 0 0 1 1 . NO
p a r t o f t h i s p u b l i c a t i o n m a y be r e p r o d u c e d , s t o r e d in a r e t r i e v a l s y s t e m , o r t r a n s m i t t e d , in
a n y f o r m , ; o r bAyllfany, fm~ainS~neleCtrmOn~C~nmn~h~2~akl p h o t o c o p y i n c j , m i c r o f i l m i n g . . . . . . ding,
112 Lovaas, Schreibman, and Koegel

From a behavioristic viewpoint, it is quite unnecesary to postulate such an

underlying disease or entity, and indeed it is quite possible that the different
autistic behaviors are related to several different kinds of antecedent conditions.
For example, we do know that many if not all of the behaviors one observes in
autistic children exist in other children as well; retarded and blind children
self-stimulate as do normal children if they have nothing else to do. Brain-
damaged children sometimes echo while some retarded children have unusually
well-developed rote memories, and so on.
In our treatment we have felt that we could develop procedures to help
these children overcome their pathological behaviors and develop healthy ones
without having to postulate an underlying process such as autism. Instead, we
have thought that we may be able to isolate the controlling conditions for each
one of their various pathological behaviors, taken one at a time. It would of
course be desirable, if for no other reason than treatment efficiency, if we could
discover that their various behaviors interacted in the sense that if we changed
one behavior, then certain others would change concurrently. But one may also
be prepared for the possibility that these behaviors are relatively independent of
each other so that as one gains some control over one of them, one does not
necessarily gain control over the others. For example, as one established the kind
of relationship with a child which makes him affectionate to adults, he might
not simultaneously show any improvement in his language. However, should a
speech therapist be successful in teaching language to such a child, then one
might observe a concurrent decrease in psychotic speech. Needless to say, we
hope for large interacting response classes, since our treatment efforts would
then become proportionately more efficient.
Given the uncertainty both in regard to etiology and prognosis of the
diagnostic label autism, we would prefer to provide a diagnosis of the child in
terms of the specific behaviors he does or does not have. Such a behavioral
diagnosis may be productive for several reasons. First, if the variables which
control a particular behavior are known, then the treatment is suggested by the
diagnosis. If we do not know how to alter the behavior, then the diagnosis would
suggest that further research need be attempted before these behaviors can be
treated. Therefore, a prognosis is also included in the behavioral diagnosis.
Finally, describing a child in terms of specific behaviors, since these are relatively
public and visible to all, should help to facilitate communication between those
who produce research and those who consume it.

BEHAVIOR THEORY AND AUTISM

The first succinct attempt to understand the behavior of autistic children

within a behavioristic framework was carried out by Ferster (1961). Ferster
A Beha~or Modification Approach 113

presented a very convincing argument of how it was that in the absence of

acquired symbolic rewarding aspects of social stimuli (and with a general defi-
ciency in acquired reinforcers), one might expect the very impoverished behav-
ioral development one sees in autistic children. The primary contribution of
Ferster's theoretical argument lies in the explicitness and concreteness with
which he relates learning principles to behavioral development. There have been
some general efforts to relate learning theory to psychopathology in the past,
but perhaps none presented the argument as directly as did Ferster.
Shortly thereafter, Ferster and DeMyer (1962) reported a set of experi-
ments in which they exposed autistic children to very simplified but controlled
environments where they could engage in simple behaviors such as pulling levers
or matching to sample, for reinforcers which were significant or functional to
them. The Ferster and DeMyer studies were the first to show that the behavior
of autistic children could be related in a lawful manner to certain explicit envi-
ronmental changes. What the children learned in these studies was not of much
practical significance, but the studies did show that by carefully programming
certain environmental consequences, these children could in fact be taught to
comply with certain aspects of reality.
These early studies and others had certain features in common. They
explicitly arranged certain relationships between their patient's behavior and his
surroundings and immediate environment. The data which emerged fronl these
studies were on the whole very regular and lawful, which is another way to say
that one could understand the patient and be of help to him. Finally, because
the majority of the patients whom these investigators worked with had been so
difficult to help before, a wave of optimism and enthusiasm was communicated
about what might be done for severely disturbed, such as psychotic and retarded
children. It seemed, in these early studies, that the problem which faced us could
be solved if we paid attention to perfecting our educational techniques for
teaching appropriate behaviors, rather than making research efforts into some
hypothetical entity within the child, such as brain damage or psychosis.
let us now turn to a very brief summary of the essential points within that
part of learning theory which we have relied upon to help us in our treatment
projects with psychotic children. Using the concepts of learning theory, one can
view a child's development as consisting of the acquisition of two events: (a)
behaviors and (b) stimulus functions. If we look at the behavioral development
of autistic children, perhaps the most striking feature about them centers on
their behavioral deficiency. They have little if any behavior which would help
them function in society. If one was going to treat autistic children based on this
perspective, then one would try to strengthen behaviors, such asappropriate play
and speech, by reinforcing their occurrence. When their occurrence is initially
absent, those behaviors should be gradually shaped by rewarding successive
approximations to their eventual occurrence. Similarly, one might attempt to
treat certain behaviors, such as tantrums and self-destruction, by either system-
114 Lovaas, Schreibman, and Koegel

aticaUy withholding those reinforcers which may be maintaining these behaviors

or by the systematic application of aversive stimuli contingent on their occur-
rence. In other words, it would be possible to develop a treatment program
where one worked directly with the child's behaviors, using whatever reinforcers
were functional for that child. In the programs we have developed so far, we
have usually developed behaviors through primary reinforcers such as food.
The child not only acquires behaviors as he develops, but his environment
also acquires meanings (stimulus functions) for him. On part of the meaning that
the world has for a person centers on its perceived reward and punishment
attributes. We speak here of symbolic rewards and punishments, which techni-
cally are referred to as secondary or conditioned reinforcers. That is, certain
parts of the child's environment which were neutral when he was born come to
acquire the function of rewarding and punishing him. One can think of many
good examples to illustrate this point, and it is particularly obvious that within
the social area much of this kind of learning takes place. The presence or
absence of an approving smile, while neutral to the newborn infant, gradually
assumes reinforcing functions as the child interacts with his parents. The primary
reason that the acquisition of these secondary reinforcers is so important lies in
their control over the acquisition of behavior. Normal children appear to acquire
much of their behaviors on the basis of secondary reinforcers. If autistic children
do not respond to or become minimally affected by praise, smiles, hugs, inter-
personal closeness, correctness, novelty, and other such secondary reinforcers
which support so much behavior in normal children, it would be logical to argue
that their behavioral development should be accordingly deficient. Thus, much
of an autistic child's failure to develop appropriate behavior could be viewed as a
function of a more basic failure of his environment to acquire meaning for him;
that is, his environment failed to acquire secondary reinforcers. This was the
essence of Ferster's (1961) theoretical analysis of autism.
Since these early papers, several studies have applied behavior modification
procedures in an attempt to treat or educate autistic children. These studies have
been characterized by relatively sophisticated research designs, which have
allowed the investigators to draw conclusions about the effectiveness of their
interventions; most of the potential thrust in behavior modification research
derives from this adherence to sound research designs. Most o f the behavioral
studies have relied upon single-subject (or within-subject) replication designs, so
that one could be reasonably certain that the treatment which was given did in
fact help that child. But since the studies were limited to single children, the
effect of the treatment across several children was not known. In one of the first
and better known studies, Wolf, Risley, and Mees (1964) used reinforcement
principles and a reversal design to treat a 31/.,-year-old autistic boy who would
not eat properly, lacked norlnal social and verbal repertoires, was self-destruc-
tive, and refused to wear glasses necessary to preserve his vision. The child's
A Behavior Modification Approach 115

behavior improved markedly and a subsequent follow-up study (Wolf, Risley,

Johnston, Harris, & Allen, 1967) showed that he continued to improve after
discharge from the program, to the point where he was able to take advantage of
a public school education program.
In another early study, Hewett (1965) described a procedure for building
speech in children who were initially mute. He used a shaping procedure to
increase the child's attending behavior, then systematically rewarded the child
for vocalizations that eventually matched those modeled by the therapist. Using
this procedure the child acquired the beginnings of meaningful language.
What the studies of Wolf, Risley, and Mees (1964) and of Hewett (1965)
showed was that it was quite possible to take learning-theory principles which
had been discovered in a laboratory setting, to move outside the laboratory and
into the child's day-to-day environment, and to begin to work with behaviors
which were directly and clinically relevant, such as to reduce tantrums and
atavistic behavior and to establish meaningful social behaviors. These studies
relied on what was known already within learning theory to carry out their
therapeutic programs. Their use of extinction, shaping by successive approxima-
tions, and similar principles had all been isolated in animal research.
When one works with children who have severe behavioral deficiencies, it
soon becomes apparent that there are some real limitations in the degree to
which one can build complex behavior repertoires into children by using direct
shaping procedures. Observations of the development of normal children sug-
gested that the acquisition of complex behaviors was facilitated through imita-
tion of such behaviors in adults and peers (Bandura, 1969). In work with autistic
children, it was soon discovered that these children were greatly deficient in
imitative behavior and seemed to learn little if anything on an observational
basis. Therefore, to make significant progress in the acquisition of complex
repertoires, we needed to know more about the conditions under which imita-
tion occurred. Fortunately, Baer and Sherman (1964) published a very impor-
tant study which showed that if one reinforced a child for imitating some of a
model's behavior, the child would also begin to imitate other behaviors of the
model, even though he had not been explicitly reinforced for imitating these.
They viewed imitation as a discrimination, a situation in which the child discrim-
inates the similarity between his and the model's behavior as the occasion for
reinforcement. Although the Baer and Sherman study dealt with normal children
who already imitated, its results gave rise to procedures for building imitative
behavior in nonimitating children. Thus, Metz (1965), using reinforcement-theory
principles, presented the first study to show how one could use reinforcement
theory principles to build nonverbal imitative behavior in nonimitating autistic
children. Lovaas, Berberich, Perloff, and Schaeffer (1966) showed how it was
possible, through the use of a discrimination-learning paradigm, to build imitative
verbal behavior in previously mute autistic children.
116 Lovaas, Schreibman, and Koegel

In reviewing the history of this development, it is apparent that barely

twelve years have elapsed since Ferster (1961) published the first theoretical
article which attempted a behavioral analysis of autism and that not more than
nine years have elapsed since the first treatment study was published. Obviously
this is a very new field that has experienced a promising start.

RECENT DEVELOPMENTS

We will now go on to illustrate some of the more recent developments

within the behavior modification approach to autistic children. As we do so it
will become apparent that a behavioral approach to autism has not addressed
itself to the total child or autism as an entity, whatever that means. Neither has
this behavioristic approach committed itself to a particular etiology of autism. It
is important to point out that it would be very significant to discover whether
there is a phenomenon called autism and to discover a specific etiology o f such a
problem. However, the search for answers to such questions is severely restricted
because of limitations on research methodology. So, for the time being, we
refrain from posing questions to which it seems impossible to obtain answers
today. On the other hand, the children present certain immediate problems
which we can attempt to ameliorate on the basis of what we know today.

Analysis of Self-Destructive Behavior

One of the most bizarre and most profoundly sick behaviors which one
will ever encounter is self-mutilation, which is characteristic of so many psychot-
ic and retarded children. One can see children who tear with their teeth large
amounts of tissue from their shoulders and arms, who chew off part of their
fingers, hit their heads so violently against the wall that they detach their retinas,
and accidentally would kill themselves unless they were somehow sedated or
restrained. Many of these children, although they may be only eight or nine
years old, have spent most of their lives in restraints, tied down both by their
feet and arms.
Irrational as these behaviors may appear, the studies which were con-
ducted to better understand these behaviors show self-mutilation to be a very
"lawful," understandable phenomenon. Rather than the self-destruction being
an expression of some tenuous internal state, such as a shattered, guilty, worth-
less self, we found it to be rather straightforward learned behavior. That is, we
first attempted to treat self-destruction using the treatment procedure most
often prescribed at the time. This treatment, based on the psychodynamic model
of psychopathology, suggested delivering affection, understanding, and sympa-
thetic comments to reassure the child of his self-worth, etc. We carefully mea-
A Behavior Modification Appzoach 117

I / m

j ~

4j

r
uJ
o3
z

5 -~ 2 I/2 MIN

Fig. 1. Greg's self-destruction, as cumulative response

curves, over successive sessions (1 through 7). The upward-
moving hatchmarks in Sessions 3 and 5 mark delivery of
sympathetic comments, play, etc., contingent on self-de-
struction.

sured the amount of self-destructive behavior emitted by the child and found
that when we gave sympathetic comments contingent upon the child hurting
himself, his self-destruction increased in strength, only to return to baseline
when these comments were no longer administered. Figure 1 presents the data
for one child, Gregory, indicating the change in his self-abusive behavior as a
function of the attention he received for such behavior. His data are presented as
cumulative curves. This means that if he did not hit himself, the line is flat
(i.e., horizontal); and each time he injured himself, it is recorded on a pen that
makes one small step upward for each response. The steeper the curve, the more
hits. Notice that we have in sessions 1 and 2, baseline data that tell us that his
118 Lovaas, Schreibman, and Koegei

self-destructive behavior is very low; the curves are almost flat. In session 3, we
started our treatment with him, expressing our concern and affection for him
when he hit himself. Note also that each time we treated him in this manner the
pen on the graph gave a signal by an upward-moving stroke or hatchmark. In
session 3, he showed a marked increase in the rate of self-destruction when we
treated him like that. In session 4, we removed the treatment and the curve
flattened out; he got better. In session 5 we reintroduced the treatment and he
hit himself some 200 times in the short period of 10 minutes. When we again
removed the treatment (sessions 6 and 7) we recovered his baseline. This, then, is
an ABABA design. Incidentally, it would have been quite possible for us to have
killed Greg through this treatment. All we needed to do was to thin out the
reinforcement schedule a bit (i.e., follow self-destructive acts by sympathetic
comments less than 100 percent of the time) and to have given love contingent
on more vicious blows, closer to the eyes or other vulnerable spots. In other
words, it looked like the self-mutilation was reinforceable, like operant behavior.
It is noteworthy that this is the first instance reported in the literature where
objective measures were taken to carefully assess the effects of treatment for
self-destruction. Clearly the emphasis on a research framework has enabled us to
see quite vividly the dangerous prospects of this form of treatment.
We hypothesized that if self-destruction were reinforceable, as operant
behavior is, then it should e x t i n ~ i s h if we made certain that no social conse-
quences were administered contingent upon its occurrence. This is exactly what
happened. When the children were left free to injure themselves without parents
or nursing staff intervening, the self-destructive behavior fell off in a very gradual
but lawful manner from, for example, a high of some 4000 self-destructive acts
in the first hour to 3500 the second hour, then 300, and then slowly to zero by
the tenth session. What most convinced us that self-mutilation was in fact
operant behavior pertained to its highly discriminated nature. For example, the
child would not hurt himself in a room where he had been "run to extinction"
but would resume self-destruction at full strength in another situation only feet
and seconds away if he had been comforted for self-destructive behavior there.
Clinically speaking, the children did not waste any blows unless there was a
payoff for it, and our data showed that they were incredibly discriminating as to
which situations paid off. Technically speaking we say that self-mutilation was
under the control of discriminative stimuli (SD).
Although extinction may work to reduce self-destruction, there are serious
limitations to extinction procedure since some of the children hurt themselves
severely during extinction. Clearly, the child who hits himself thousands of times
in an hour's time, or a child who tears flesh off his shoulder, is in danger, and
one takes too many chances letting such a child undergo extinction runs as we
have described them. Therefore, we tried out ways of delivering aversive stimuli
(such as painful electric shock) contingent upon self-destruction, in the hope of
stopping it more quickly. The data we reported (Lovaas & Simmons, 1969)
A Behavior Modification Approach 119

showed an immediate and very dramatic termination of the self-destructive

behavior, even in children who had been self-destructive for years. It is unclear
why a relatively innocent aversive stimulus should terminate self-destruction,
considering the very severe physical abuse the children inflicted upon them-
selves. It is possible that the children had adapted themselves to the pain from
their self-inflicted injuries, while the electric shock was new, offering no oppor-
tunity for adaptation. Several people have now published on the control of
self-destructive behavior (Bucher & Lovaas, 1968; Risley, 1968; T a t e & Baroff,
1966) and the data are remarkably consistent.
It is likely that we understand self-destructive behavior better than any
other of the behaviors which psychotic children bring to treatment. It is also
significant to point out that the analysis we have presented for self-destructive
behavior is the kind of analysis which behavior modifiers undertake; that is,
trying to understand one behavior at a time, identifying the conditions which
control the behavior, and observing how changes in one behavior may alter
others. Surprisingly, and disappointingly, the reduction in self-destructive behav-
ior did not bring with it a simultaneous change in large classes of other behav-
iors. The children who stopped mutilating themselves did not simultaneously
become normal.

SelfiStimulatory Behavior
By far the most common form of behavior exhibited by psychotic and
retarded children centers on the self-stimulatory behavior we mentioned earlier.
This includes a great deal of spinning, twirling, rocking, and gazing as well as
other very repetitive and stereotyped movements which seem to have no partic-
ular relationship to what is happening in the child's day-to-day life.
Our understanding of this behavior class is very limited. From clinical
observation, and not so much based on systematic research, it appears that this
behavior disappears or is supplanted when more nonnal behaviors are acquired.
We do know also that the presence of certain forms of self-stimulation makes the
child more difficult to teach, apparently because he is less attentive to external
cues. For example, in a study by Lovaas, Litrownik, and Mann (1971) autistic
children were taught to respond to an auditory input in order to obtain candies
and other sweets. If the child was presented with this auditory stinmlus when he
was involved in self-stimulatory behavior, then frequently he would fail to
respond to the auditory input, either completely missing the opportunity to
obtain his reinforcer or responding after some delay. In another study by Koegel
and Covert (1972) autistic children were taught a very simple discrimination task
such as responding on a machine (depressing a lever) in the presence of a light
stimulus. They observed that the children failed to acquire this very simple
discrinlination as long as they were allowed to self-stimulate. When given disap-
proval (either verbal or with a slap on the hand) resulting in the suppression of
120 Lovaas, Schreibman, and Koegel

self-stimulation, the children did learn. It appears, then, that the self-stimulatory
behavior may interfere with one's efforts to teach these children; for the sake of
helping them acquire new behaviors one may initially decrease self-stimulatory
behavior.
Recent research in our lab, largely under the direction of Arnold Rincover
and Crighton Newsom, has proceeded on the notion that self-stimulatory
behavior is operant behavior and that it is different from other kinds of operants
in the sense that the reinforcement for such behavior consists of the sensory
feedback (kinesthetic, visual, auditory, etc.) which the behavior itself produces.
In that sense it is the child himself who controls that reinforcement, hence
shaping his own behavior. Given the extensive and durable self-stimulatory reper-
toires of autistic children, we suspect that these self-stimulatory reinforcers must
be very strong, and we are exploring ways of using self-stimulatory reinforcers
instead of food, etc., to build appropriate behaviors. So far research shows
self-stimulatory reinforcers to be much less satiable than the primary, appetitive
ones.

The Teaching of Appropriate Behaviors

Since these children have such limited repertoires and since many are
essentially without any social or intellectual behavior, they present a great
challenge to educators. In a sense, one has the opportunity to start building a
person from the beginning in regard to social and emotional development as well
as intellectual development. In the short amount of space allotted here we can
only point to certain examples to illustrate the techniques which have been
employed in our attempts to build new behaviors.
Simultaneously with the suppression of self-stimulatory behaviors, the
teacher generally begins her work by attempting to establish some early forms of
stimulus control. The teacher may request of the child some simple behavior
such as sitting quietly in a chair. Since even such a minimal request often evokes
tantrums of self-destructive behavior, the establishment of this basic stimulus
control and the reduction of aggression and self-destructive behaviors generally
proceed together. It is generally impossible to work on the acquisition of
appropriate behaviors until one has achieved some reduction of the pathological
behaviors.
When one attempts to educate children who are so deficient in behavioral
development, theoretically one has two alternate paths to follow. For example,
instead of building behaviors piece by piece, one could decide to teach a child to
value interpersonal interactions, intellectual achievement, curiosity about the
world, and so on. Technically speaking, one could attempt to place these chil-
dren in a situation where one optimized the child's motivational system to help
them acquire conditioned reinforcers. We have pointed out before (Lovaas,
A Behavior Modification Approach 121

Freitag, Kinder, Ruberstein, Schaeffer, & Simmons, 1966) that in many ways
this would be an optimal strategy because the child, being properly motivated,
would learn much without explicit attempts to teach him. To illustrate this
point, if a child were strongly reinforced by a large range of social reinforcers, he
would probably learn to behave in social ways, that is, to talk, play with peers,
and so on, so as to come in contact with these stimulus events. It is unfortunate
that this alternate path to teaching is not open to us since we do not know how
to establish conditioned reinforcers or to build motivation in autistic children.
No doubt one of the great challenges in the years ahead will be to achieve a
better understanding of how motivation is acquired.
Given this limitation on the child's motivational system, the alternative is
to proceed to build behaviors with the reinforcers which are functional. One can
always fall back upon basic reinforcers which are functional. One can always fall
back upon basic reinforcers such as food and pain, and as one works with any
one child, one discovers idiosyncracies in his motivational structure, such as a
particular liking for a certain piece of music, a toy, etc., which can be parti-
tioned in various ways and delivered to the child contingent upon certain desir-
able behaviors. The acquisition of new behaviors is accomplished in a step-by-
step program. Let us turn to eftbrts to build language to ilhistrate how these
programs have worked out.

Building Language
Using a shaping procedure, several investigators (Hewett, 1965; Lovaas,
1966) have provided procedures for developing speech in previously mute
autistic children. The procedure relied heavily on developing imitative speech.
Let us illustrate the procedures from our language program. This program has
been presented on f'dm (Lovaas, 1969) and will appear in a book (Lovaas, in
press). Briefly described, the verbal imitation training involves four steps of
successive discriminations. In Step l the therapist increases the child's vocaliza-
tions by reinforcing him (usually with food) for such behavior. In Step 2 the
child's vocalizations are reinforced only if they are in response to the therapist's
speech (e.g., if they occur within five seconds of the therapist's speech) until he
can match the particular sound given by the therapist (e.g., "a"). In Step 4 the
therapist replicates Step 3 with another sound (e.g., "m"), demanding increas-
ingly f'me discriminations and reproduction from the child. In this manner,
starting with sounds which are discriminably different (such as " m " and "a"),
the child is taught to imitate an increasingly large range of sounds, words, and
sentences. Imitation, then, is a discrimination where the response resembles its
stimulus. Once he can imitate, the previously mute child becomes similar to
echolalic autistic children; they both imitate the speech of others, but neither
knows the meaning of the words he utters. Their speech exists without a con-
text.
122 Lovaas, Scltreibman, and Koegel

A program for the establishment of meaningful speech involves establish-

ing a context for speech which consists of two basic discriminations. In the first
discrimination the stimulus is nonverbal and the response is verbal, as in expres-
sive speech (e.g., the child may be taught to label a food). In the second discrim-
ination the stimulus is verbal but the response is nonverbal, as in language
comprehension (e.g., the child learns to follow instructions, obey commands,
and so forth). Most language situations involve components of both discrimina-
tions: the stimulus and response have both verbal and nonverbal components.
The speech program, based on these two discriminations, begins with simple
labeling, which is made functional as soon as possible. For example, as soon as a
child knows the label for a food, he is fed contingent upon asking for food. The
program gradually moves on to make the child increasingly proficient in lan-
guage, including training in more abstract terms (such as pronouns, time, etc.);
some grammar, such as the tenses; the use of language to please others, as in
recall or storytelling, and so forth. These later levels are only rarely reached by
mute children, but are ahnost always reached by echolalic children. Other inves-
tigators using similar procedures report similar data on language training (Risley
& Wolf, 1967).
Throughout the treatment program there is an emphasis on teaching the
child behaviors which are both socially desirable and useful to him. Thus, while
the majority of the research has focused on attempts to build language, there
have also been several attempts to facilitate social and self-help skills. Lovaas,
Freitas, Nelson, and Whalen (1967) published a procedure for building nonverbal
imitation which proved particularly useful for the purpose of developing social
and self-help skills. It includes methods for building those behaviors which make
the child easier to live with, such as friendly greetings and shows of affection,
dressing himself, feeding himself, brushing his teeth, and so on. Again the
method is based on shaping procedures where the child is rewarded for making
closer and closer approximations to the attending adult's behavior. As the chil-
dren learn to discriminate the similarity in their own and the model's behaviors,
they gradually acquire imitative behavior, as they did in the speech program.

GENERALIZATION AND FOLLOW-UP RESULTS

We now have some data which provide an estimation of the changes one
might expect in autistic children undergoing behavior therapy (Lovaas, Koegel,
Simmons, & Long, 1973). We examined three measures of the generality of
treatment effects: (a) stimulus generalization, the extent to which behavior
changes that occur in the treatment environment transfer to situations outside
that treatment; (b) response generalization, the extent to which changes in a
limited set of behaviors effect changes in a larger range of behaviors; and (c)
generalization over time (or durability), that is, how well the therapeutic effects
maintain themselves over time.
 A Behavior Modification Approach 123

Let us illustrate the kinds of treatment changes and follow-up data we

have collected by presenting certain data on the first ten children we treated. We
recorded five behaviors in a free-play situation which was different from the
treatment environment, and in the presence of people who had not treated the
child. Two of these behaviors were "sick" behaviors--self-stimulation and
echolalia, which we have described earlier. Three of the behaviors were
"healthy" -- appropriate verbal behavior, which was speech, related to an appro-
priate context, understandable, and gramatically correct; social nonverbal, which
referred to appropriate nonverbal behavior that depended on cues given by
another person for its initiation or completion; and appropriate play, which
referred to the use of toys and objects in an appropriate, age-related manner.
The recordings were made before treatment started, at the end of treatment
(after 12 to 14 months), and in a follow-up some one to four years after treat-
ment. The children were divided into two groups - those who were discharged to
a state hospital and those who remained with their parents.
The data are presented in Figure 2. Percent occurrence of the various
behaviors is plotted on the ordinate for before B and after A treatment and
shows the latest follow-up Fmeasures. 1 refers to the average results for the four
children who were institutionalized (discharged to a state hospital), and P refers

60
~_~stitution

50 I
I

L~J
l
I
(.~
!
zu J 40 I
OE i
n,-

0
30 l
i
I--- I
Z
~ 2o ,' ~)
13..

10 ~) "k~arents

I I I I I I l t I I I I L J J
B A F B A f B A F B A F B A F
Self Echololio Appropriate Social Appropriate
Stimulation Verbal Nonverbal PIo y
Fig. 2. Multiple response follow-up measures. Percent occurrence of the various behaviors is
plotted on the ordinate for Before (B) and After (A) treatment, and for the latest follow-up
(F) measures. 1 refers to the average results for the four children who were institutionalized,
and P refers to the average results for the nine children who were discharged to their
parents' care.
124 Lovaas, Sehreibman, and Koegel

to the six children who lived with their parents after their discharge from treat-
ment. For all the five behaviors the trends are the same: the children who were
discharged to a state hospital lost what they had gained in treatment with us: they
increased in their psychotic behavior (self-stimulation and echolalia), they appear
to have lost all they had gained of social nonverbal behavior, and they lost much
of their gains in the area of appropriate verbal and appropriate play. The children
who stayed with their parents, on the other hand, maintained their gains or im-
proved further. For the children who regressed in the state hospital, a brief re-
instatement of behavior therapy could temporarily reestablish the original
therapeutic gains.
Since we assessed these behaviors in environments other than that of the
treatment, we know that our procedures produced stimulus generalization. The
children's Stanford-Binet IQ scores and Vineland Social Quotient scores also
showed large gains during the course of treatment, and since we did not train the
children on these tests, they provide explicit measures of response generaliza-
tion.
These findings clearly emphasize an important point underlying the use of
principles of behavior modification. That is, there may be important differences
in the procedures for the production and maintenance of behavior. Thus, it does
not appear to be enough to help the child acquire appropriate behaviors and to
overcome the inappropriate ones; it is also important to provide maintaining
conditions which ensure that the improvements will last. Our follow-up paper
(Lovaas et al., 1973) discusses several of the strengths and weaknesses of a
behavior modification approach to the treatment of autistic children.

CONTEMPORARY RESEARCH

Behavior therapy for severely psychotic children requires extensive teacher

involvement; it is incremental and slow and the therapeutic gains are reversible.
Many very important questions still remain unanswered. Why do certain autistic
children show much larger improvements than others? Why does a given autistic
child show relatively large and rapid improvements in some areas and slow,
minimal gains in other areas? Why do so few autistic children become normal?
Can autistic children function in public schools? Let us conclude this chapter by
reviewing some current research which will hopefully suggest improvements in
the use of behavior modification with autistic children.
Many researchers and clinicians have emphasized the extreme inconsis-
tency with which autistic children respond to sensory input. At one time they
appear to be blind and deaf, while at another time they show extremely fine
visual and auditory acuity. In one situation they respond correctly to an instruc-
tion, while at another time they appear to have learned nothing about how to
A Behavior Modification Approach 125

respond. Such peculiarities in the children's responding led us to conduct the

following studies.
In the first study (Lovaas, Schreibman, Koegel, & Rehm, 1971) we trained
normal children and autistic children to respond to a complex stimulus involving
tile simultaneous presentation of auditory, visual, and tactile cues. Once this
discrimination was established, elements of the complex were presented sepa-
rately to assess which aspects of the complex stimulus had acquired control over
tile child's behavior. We found that the autistic children had primarily come
under the control of only one of the cues; tile normal children responded uni-
formly to all three cues. We also found that we could arrange conditions such
that a cue which had remained nonfunctional when presented in association with
other cues could be established as functional when trained separately; thus, the
autistic children did not appear to show a deficit in any particular sensory
modality. Rather, when presented with multiple sensory input, a restricted range
of that input gained control over their behavior.
In a teaching situation, when tire child can both see tile teacher's lace as
well as hear him talk. tile autistic child may solve the verbal imitation problem
by attending only to tile teacher's face. We referred to this finding as stimulus
overselectiviO, and pointed out that it had many implications for understanding
tile behavior and learning problems of autistic children. For example, if autistic
children are generally overselective in their response to multiple cues, they may
be functionally blind in those situations where they are "hooked" on auditory
cues and functionally deaf in those situations where they are " b o o k e d " oil visual
cues. In a related study (Schreibman & Lovaas, 1973), autistic children were
trained to discriminate between a boy doll and a girl doll. The results showed
that for some children this discrimination would diminish to chance level when a
specific stimulus (such as tile shoes) was removed from the dolls, pointing out
tile extreme selectivity with which autistic children respond to their environ-
ment. Such stimulus overselectivity also has implications for understanding why
autistic children learn certain tasks so slowly. A necessary condition for much
learning involves a contiguous or near contiguous presentation of two or inore
stimuli. If autistic children do not respond to one of tile stimuli, certain acquisi-
tions may then fail to occur; for example, their acquisition of affect may be
retarded, as well as their development of meaningful speech. Tile establishment
of meaningful speech involves establishing a context for speech and thus requires
response to multiple inputs. For example, in attempting to teach a child to say
the word br the most common training procedure {"this is a book") involves
a cross-modality shift. If the clfild responds to the auditory input, he may fail to
perceive the visual referent and hence not associate tile appropriate label.
We now have data which bear directly on this problem of shifting stimulus
control. These data show that autistic children often selectively respond to
prompt stimuli to the exclusion of training stimuli (Koegel, 1971). Two groups
126 Lovaas, Schreibman, and Koegel

of children, autistic and normal, were pretrained on a color discrimination task.

Later, the colors were presented simultaneously with training stimuli (e.g., two
geometric forms, two tones) in a prompt-fading procedure. Gradually fading the
color prompt generally produced acquisition of the training discrimination for
normal Ss but not for autistic Ss. The autistic Ss continued to selectively re-
spond to the prompt until it was entirely faded out, at which point they began
to respond at chance level. They had learned nothing about the training stimuli
(e.g., forms). Surprisingly, the autistic Ss acquired these same training discrimi-
nations when prompts were not used. That is, the usual technique of providing
the children with extra stimuli to guide their learning may be exactly what
makes it so difficult for them to learn.
It is interesting to note that while the autistic children generally failed to
transfer from the color prompt to the training stimuli, they became very good at
discriminating minute differences in the color prompts. That is, perhaps stimulus
overselectivity may have benefited the autistic children's acquisition of the very
difficult faded-color discrimination. Once trained on the color cue in the initial
pretraining, stimulus overselectivity may have functioned to decrease the chil-
dren's acquisition of discriminations along continuums other than color but to
increase their acquisition of discriminations along the color continuum. Thus, it
might be a better procedure to use a training technique such as transfer along a
continuum (Lawrence, 1952) which would actually take advantage of an autistic
child's overselectivity. A study from our lab (Schreibman, 1972) examined this
hypothesis. Results showed that when the prompt stimuli fell along the same
continuum as the training stimuli, prompting facilitated the acquisition of a very
difficult discrimination which the children showed no evidence of acquiring,
either without a prompt or with a prompt, in a continuum other than the
training stimuli.

CLASSROOM TREATMENT

Little if any systematic research has been carried out on the education of
autistic children in classroom settings. Research to date has focused primarily on
the treatment of autistic children in a one-to-one teacher-child ratio. Economi-
cally, this makes the treatment unfeasible in many hospital and school situa-
tions. Therefore, we recently began a research program concerned with the
teaching of autistic children in groups (Koegel & Rincover, 1974). Let us illus-
trate this research by citing a study where the emphasis is on establishing a
teacher as a discriminative stimulus (SD) for appropriate behavior even though
the child is in a large group of children. To achieve this end we began working
with two children, using one teacher and two teacher aides. The teacher provides
commands and instructions and the aides deliver contingent rewards and punish-
A Behavior Modification Approach 127

ments. As the children become more and more proficient, reinforcement for
appropriate behaviors becomes increasingly intermittent and additional children
are introduced into the group, one at a time. If behaviors deteriorate at any one
time, the size of the class is immediately reduced, reinforcements become more
dense, and shortly thereafter we begin building again. Through a process like
this, then, it is hoped that the autistic child will be able to behave appropriately
in a more average (normal) classroom.

NEW DIRECTIONS

The systematic manner in which we have evaluated our treatment proce-

dures has allowed us to carefully plan the future directions o f our work in some
detail. We have varied the age of the children we have treated, and the data
clearly show that the younger the autistic child, the more progress he will make
in treatment. In fact, we treated four children who were under three years of
age. Three of the four children made substantial gains and at this time appear
headed toward normalcy. We are currently involved in a large-scale operation
testing age as a variable in treatment.
Another important implication o f the follow-up work is that in order to
provide for the maintenance and generalization of treatment gains, it is necessary
to provide for the continuation o f treatment outside the clinic. Thus, our main
emphasis now is on training the parents in the application of the behavior
therapy techniques. We have developed a systematic, detailed, and intensive
program for training parents of autistic children to become behavior modifiers
(Koegel, Schreibman, & Lovaas, 1973). To place the therapy in the hands of the
parents assures generalization of the treatment effects to the child's natural en-
vironment and helps maintain the treatment gains.
Finally, we have to design teaching procedures which are more efficient
than the ones we now have. We are now very careful in our use of prompt and
attempt to use those prompts that will facilitate rather than interfere with
learning. Thus we avoid using prompts that require the child to attend to
multiple cues and instead use prompts that fall within the same dimension of the
training stimulus. For example, this may involve exaggerating the relevant
stimulus of a discrimination (or presenting it alone) and gradually fading in the
irrelevant or incorrect stimulus; the child need respond to only the relevant
stimulus throughout training.

SUMMARY COMMENT

We know more about some behaviors than others. For example, we know
more about self-destructive behavior than self-stimulation. We were of more help
128 Lovaas, Schreibman, and Koegel

to some children than to others. If the child possessed a verbal topography at the
beginning o f treatment, even if it was socially nonfunctional (such as echolalia),
then we could help him substantially in developing a meaningful language. But
we were less proficient in establishing new behavioral topographies in autistic
children when none existed. For example, if the child was mute, his progress
with behavior modification procedures was very limited. Follow-up data showed
that in order to maintain the gains which the child made in treatment, he had to
remain within an extension o f the therapeutic environment. For example, for
the c l ~ d to continue showing improvement, his parents had to be taught behav-
ior modification principles. The delivery of contingent, functional reinforcers
has repeatedly been demonstrated to be a most significant feature b o t h in
providing for learning and in maintaining the behavioral gains.
As we worked with these children, it became more apparent that they
possessed certain deviations in perceptual functioning, particularly in responding
to multiple stimulus inputs, which necessitate revisions o f the usual manner o f
presenting educational m a t e r i a l Only additional research will enable us to struc-
ture such optimal learning environments.
The f r s t 12 years in behavior m o d i f i c a t i o n relied heavily on learning
principles derived from the animal laboratory and their ready application re-
sulted in rapid progress. It is likely that the next ten years will demand consider-
able new research directed specifically to the autistic child in order to help him
develop appreciably further.

REFERENCES

Baer, D. M., & Sherman, J. Reinforcement of generalized limitation in young children.

Journal of Experimental Child Psychology, 1964, 1, 37-39.
Bandura, A. Principles of Behavior Modification. New York. Holt, Rinehart, & Winston,
1969.
Bucher, B., & Lovaas, O. I. Use of aversive stimulation in behavior modification. In M. R.
Jones (Ed.), Miami symposium on the prediction of behavior 1967: Aversive stimu-
lation. Coral Gables. University of Miami Press, 1968.
Ferster, C. B. Positive reinforcement and behavioral deficits of autistic children. Child
Development, 1961, 32, 437- 456.
Ferster, C. B., & DeMyer, M. A method for the experimental analysis of the behavior of
autistic children. American Journal of Orthopsychiatry, 1962, 32, 89--98.
Hewett, J. M. Teaching speech to an autistic child through operant conditioning. American
Journal of Orthopsychiatry, 1965, 35, 927-936.
Koegel, R. L. Selective attention to prompt stimuli by autistic and normal children. Un-
published doctoral dissertation. University of California. Los Angeles, 1972.
Koegel, R. L., & Covert, A. The relationship of self-stimulation to learning in autistic
children. Journal of Applied Behavior Analysis, 1972, 5, 381 - 387.
Koegel, R. L., & Rincover, A. Treatment of psychotic children in the classroom environ-
ment. I. Leaxning in a large group. Journal of Applied Behavior Analysis, 1972,
Koegel, R. L., Schreibman, L., & Lovaas, O. 1. A manual for training parents in behavior
modification with atttistic children. Technical article, Institute for Applied Be-
havioral Sc'ience, University of California, Santa Barbara, 1973.
A Behavior Modification Approach 129

Lawrence, D. H. The transfer of a discrimination along a continuum. Journal of Compara-

tive and Physiological Psychology, 1952, 45, 511-516.
Lovaas, O. I. A program for the establishment of speech in psychotic children. In J. Wing
(Ed.), Childhood autism. London: Pergamon Press, 1966.
Lovaas, O. I. Behavior modification: Teaching language to psychotic children. (Instructional
film, 45 min., 16 ram. sound) New York. Appleton-Century-Crofts, 1969.
Lovaas, O. I. Teaching language to psychotic children. UnpubLished manuscript, University
of California, Los Angeles, 1972.
Lovaas, O. I., Berberich, J. P., Perloff, B. F., & Schaeffer, B. Acquisition of imitative speech
by schizophrenic children. Science, 1966, 151, 705-707.
Lovaas, O. I., Freitag, G., Kinder, M. I., Ruberstein, B. D., Schaeffer, B., & Simmons, J. Q.
Establishment of social reinforcers in two schizophrenic children on the basis of
food. Journal of Experimental Child Psychology, 1966, 4, 109-125.
Lovaas, O. I., Freitas, L, Nelson, K., & Whalen, C. The establishment of imitation and its use
for the development of complex behavior in schizophrenic children. Behaviour
Research and Therapy, 1967, 5, 171-181.
Lovaas, O. 1., Kocgel, R. L., Simmons, J. Q., & Long, J. Some generalization and follow-
up measure on autistic children in behavior therapy. Journal of Applied Behavior
Analysis, 1973, 6, 131-166.
Lovaas, O. I., Litrownik, A., & Mann, R. Response latencies to auditory stimuli in autistic
children engaged in self-stimulatory behavior. Behaviour Rescarch and Therapy,
1971, 9, 39-49.
Lovaas, O. I., Schreibman, L., Koegel, R., & Rehm, R. Selective responding by autistic
children to multiple sensory input. Journal of Abnormal Psychology, 1971, 77,
211-222.
Lovaas, O. I., & Simmons, J. Q. Manipulation of self-destruction in three retarded children.
Journal of Applied Behavior Analysis, 1969, 2, 143-157.
Metz, J. R. Conditioning generaLized imitation in autistic children. Journal of Experimental
Child Psychology, 1965, 2, 389-399.
Risley, T. The effects and side effects of punishing the autistic behaviors of a deviant child.
Journal of Applied Behavior Analysis, 1968, 1, 21-34.
Risley, T., & Wolf, M. M. Establishing functional speech in echolalic children. Behaviour
Research and Therapy, 1967, 5, 73-88.
Schreibman, L. Effects of within-stimulus and extra-stimulus prompting on discrimina-
tion learning in autistic children. Journal of Applied Behal,ior Anal),sis, in press.
Schreibman, L., & Lovaas, O. I. Overselective response to social stimuli by autistic children.
Journal of Abnormal Child Psychology, 1973, 1, (2), 152-168.
Tate, B. G., & Baroff, G. S. Aversive control of self-injurious behavior in a psychotic boy.
Behaviour Research and Therapy, 1966, 4, 281-287.
Wolf, M. M., Risley, T., Johnston, M., Harris, F., & Allen, E. Application of operant
conditioning procedures to the behavior problems of an autistic child: A follow-up
and extension. Behaviour Research and Therapy, 1967, 5, 103-111.
Wolf, M. M., Risley, T., & Mees, H. Application of operant conditioning procedures to the
behavior problems of an autistic child. Behaviour Research and Therapy, 1964, 1,
305-312.

View publication stats