THE DEVELOPMENT OF A TREATMENT-RESEARCH PROJECT FOR DEVELOPMENTALLY DISABLED AND AUTISTIC CHILDREN

0. lvAR 1.ovAAs UNIVERSilY OF CALIFORNIA, LOS ANGELES

This paper describes the development and main results over the last 30 years from the treatment­ research project with developmentally disabled (autistic) children in the Psychology Department at the Universiry of California, Los Angeles (UCLA). Three imponant dimensions in treatment research are addressed. The first penains to the role of serendipiry or accidental discoveries, the second to the imponance of pursuing inductive rather than theory-driven research, and the third to the imponance of adding in a cumulative and step-wise manner to improve treatment adequacy. Data from various areas of treatment research have been used to illustrate new directions for the project. These illustrations center on early and successful attempts to isolate experimentally the environmental variables that control self-injury, failure to observe response and stimulus genetalization with sub­ sequent loss of treatment gains, and the main results of intensive and early behavioral intervention in the child’s natural environment. Effective treatment for severe behavioral disorders is seen to require early intervention carried out during all or most of the child’s waking hours, addressing all significant behaviors in all of the child’s environments, by all significant persons, for many years.

DESCRIPTORS: inductive treatment research, generalization, early intervention, home-based treatment, children

I will take this opportunity to describe, in an informal and personal manner, the more critical observations and turning points that have helped determine new directions in treatment research with developmentally disabled and autistic children at the UCLA Psychology Department’s Autism Proj­ ect. Space does not allow for a comprehensive re­ view of the empirical studies that form the basis for behavioral treatment of developmentally dis­ abled (including autistic) children. For such reviews the reader is referred to Lovaas and Smith (1988), Newsom and Rincover (1989), and Schreibman ( 1988). One of the best known and earliest pre­ sentations of the research procedures pursued in this paper is presented in detail by Sidman (1960). Thompson (1984) has discussed certain instructive parallels between the inductive and discovety-ori­ ented research of Claude Bernard (the father of physiology) and that of applied behavior analysis. Bernard attempted to isolate helpful from harmful (medical) treatments, as is the goal of psychological treatment research. The reader may also want to

Address correspondence and reprint requests to 0. Ivar Lovaas, Psychology Department, Universiry ofCalifornia, Los Angeles, 405 Hilgard Ave., Los Angeles, CA 90024-1563.

become familiar with the work of Chamberlin ( 1897), a geologist who advocated a procedure that he labeled “the method of multiple working hy­ potheses” and warned against the seductiveness of the “grand theories” in misleading inquiry, much as is advocated in this paper. I will therefore lay no claims to originality in the observations pre­ sented here because others have discussed similar observations more succinctly.

People often ask me how I became interested in working with children diagnosed as autistic. I know that the route that took me there was not one that I designed for myself, but one that my environment arranged for me, in a rather fortuitous manner. My interest in environmental determinants stemmed from the German occupation of Norway during World War II. As a child, I wondered whether such destructive behaviors were genetically or en­ vironmentally determined. I hoped for the latter. By luck, I was assigned an adviser in the graduate program at the University ofWashington, Professor Edwin Esper, who was of “the old school” of be­ haviorists. Esper was upset with Boring’s favorable coverage of Wundt and Titchener and ended up writing his own histoty of psychology to correct Boring’s “mistakes” (Esper, 1964). He was a stu-

617

618 0. IVAR LOVAAS

student ofA. P. Weiss, had studied linguistics with Bloomfield, favored behavioral interpretations of language, was strongly critical of Chomsky, and had published one of the first experimental inves­ tigations in linguistics (Esper, 1925). Pioneers in psychology, like Thorndike, Hall, Yerkes, Meyer, Weiss, and Jastrow, were behaviorally oriented and initiated applied psychology, addressing education­ al and clinical problems. O’Donnell (1985, pp. 209-291) provides an interesting account of this development as it was welcomed by psychiatry with its ties (at that time) to biology. Behavioral psy­ chology was considered to preserve “the experi­ mental rigor oflaboratory science” (p. 235), which Freudian theory was seen to threaten. At the same time, psychologists who worked within universities were under some pressure to give help to the society that financed them (“the university … belonged to the people,” p. 217), and behavioral psychology offered a framework for studying how to change behavior. O’Donnell provides a quotation from Lighmer Witmer that is as true to behaviotal sen­ timent today as it was 90 years ago: “To ascribe a condition to the environment, is a challenge to do something for its amelioration; to ascribe it to heredity often means that we fold our hands and do nothing” (p. 234).

Given this promising start, one may wonder why it took so many years before behavioral psychology fielded large-scale treatment programs. In fact, by the 1950s, behavioral psychologists seemed antag­ onistic towards those who addressed social prob­ lems. One of the few attempts to unite behavioral and clinical psychology was presented by Dollard and Miller (1950), who translated psychoanalytic concepts and clinical observations into Hullian the­ ory. As was the case with so many conceptual e~~rts at the time, the gap between theory and empmcal observations was too large, and there were little or no data. Today, some 100 years after the pioneers expressed their hopes of providing a scientific basis for treatment and educational interventions, the gap is beginning to narrow. A major source of joy in my work has been to help narrow the distance between research and pracrice.

EARLY WORK ON LANGUAGE

I had been offered a postdoctoral position by Sid Bijou in his Child Development Institute in 1958, and was placed in an environment that eventually would help me contribute to bridging the gap be­ tween behavioral psychology and clinical applica­ tion. At the Child Development Institute, I began my professional career by introducing supervised training in psychodynamic therapies to the clinical students there (which included Robert Wahler and Ralph Wetzel). I was beginning to feel the futility of such efforts, and Sid Bijou placed me under increasing pressure to ”do research” with the pre­ school children. Not knowing what to do, I set out to test whether one could demonstrate reinforce­ ment control over the vocal utterances of preschool children, trying to replicate Greenspoon’s experi­ ments. I was not very enthusiastic about that proj­ ect, but it seemed better than doing nothing. It is noteworthy that had it not been for a very accidental discovery at that time, my research career may have been very different or nonexistent. The full story is too long to tell here (but see Lovaas, 1977, pp. 119-126). In short, a child subject was seated in a playroom in full view from where I sat behind the one-way screen in an adjoining observation room. He had been instructed to sit and then to talk to a box that dispensed toy trinkets as reinforcers. At some point he said, “What shall I say?” As per protocol, he was reinforced for making that utter­ ance. To my surprise, he then got out of his chair and walked over to the attending adult to ask this question of her. Halfway through his walk across the room, I was about to tell my assistant to “Get

the kid back in his chair,” when it dawned on me that the child was leading me to an important discovery. In short, I reasoned that the discrimi­ native stimuli (SD) generated by his immediately preceding verbal response (“What shall I say?”) exercised SD control over his own nonverbal be­ havior (walking across the room to address the assistant). He had not been instructed to get up; rather, he had been told to sit. In short, he probably would not have walked across the room, had it not

619 TREATMENT-RESEARCH PROJECT

been for his question “What shall I say?” (The immediate causes of behavior are likely to be found in the seconds preceding or accompanying a re­ sponse. It may help to think of a reflex, where the latency between a stimulus and a response is a matter of milliseconds.) It was this kind of obser­ vation that I needed to make sense out of my discussions about Benjamin Whorf (1956) with Professor Esper some 5 years earlier. Whorf had proposed that a person’s particular language may influence the way that person perceived, thought about, and behaved (nonverbally) in his or her physical environment. Similarly, Dollard and Miller (1950) had proposed that the response-produced stimuli from a client’s “new language” (at the end of treatment) would have to control new and “healthy” behaviors outside of treatment, in order for such treatment to be effective. My interest in these matters stemmed from my clinical practice. Like most clinicians, I administered ”talking ther­ apies.” My clients would ask me, “Do you mean that by sitting here talking to you, I will get better?” I answered “yes,” placing my faith on the hy­ potheses of Whorf and Dollard and Miller.

At the end of a series of studies on verbal control over nonverbal behavior (Lovaas, 1961, 1964a, 1964b), there was some evidence that, for example, if one reinforced verbally aggressive statements, then one could observe an increase in non-verbal ag­ gressive behavior. However, I could only conclude that the amount of verbal control over nonverbal behavior varied considerably across children, and when present, it seemed rather short-lived. The degree of stimulus control would vary across chil­ dren as a function of their particular reinforcement history, and could be easily extinguished. Subse­ quently, Meichenbaum was to learn of these studies (Meichenbaum & Goodman, 1969, 1971), which contributed to a foundation for what is now labeled cognitive behavior therapy. For reasons that I can­ not understand, Meichenbaum deleted the learn­ ing-based interpretations of these verbal-nonverbal interactions that I had provided. Instead, he sub­ stituted developmental and cognitive terminology, much of it from lay terminology, to account for

what was hoped to become a substantial effect in psychotherapy. Dropping the experimentally based learning interpretations may have been a mistake. A review of the literature suggests that more data are needed to support whether, and with what kind of clients, cognitive-behavioral techniques generate strong and lasting therapeutic effects (Barlow, 1988, on anxiety; Elkin, Parloff, Hadley, & Autrey, 1985, on depression; Meador & Ollendick, 1984 on con­ dua disorders and hyperactivity in children; Whit­ man, 1990, on persons with mental retardation). In his discussion of the Whorfi.an and similar hy­ potheses, Jordan (1982) points out that few hy­ potheses in anthropology, psychology, and linguis­ tics have been more appealing and, at the same time, supported by less evidence.

However limited the verbal control turned out to be, this early research helped to lead me away from research in highly contrived laboratory settings and into real-world environments. Two very sig­ nificant events took place in the late 1950s, when Ted Ayllon and Israel Goldiamond presented pro­ grams and data on how to intervene on the real­ life behaviors of schizophrenic clients and persons with speech dysfluencies. They were clinicians and researchers “in one,” presenting as good examples of reinforcement and stimulus control as anyone could want. Psychoanalytic theories of stuttering and Harry Stack Sullivan’s theory of schizophrenia were both more entertaining to me. Sadly, they were both failing to generate effective treatments. By the early 1960s, a number of studies had been published that represented the first steps in gaining experimental control over behavior in real-life en­ vironments, perhaps the most important achieve­ ments in the history of applied behavior analysis. It was a development that helped the early behav­ iorists, and later B. F. Skinner, to realize their hopes of applying experimental designs to solving social problems.

MEETING WITH AN AUTISTIC CHILD

There were a number of confounding and un­ controlled variables in the early work on the inter-

https://Whorfi.an
620 0. IVAR LOVAAS

action between verbal and nonverbal behavior. One of these was the subject’s history with language, which of course was essentially unknown. After I arrived at UCLA in 1961, my students and I went out looking for children who were chronologically of an age at which they were expected to talk, but who evidenced little if any language. Our hope was to build language in these children under controlled and known circumstances, and then to observe closely how their social, emotional, intellectual, and other behaviors might concurrently change. My at­ tempts to isolate the pivotal role of language in controlling other behaviors still had not extin­ guished.

In searching for a child who was old enough to talk but had failed to do so, my students and I were referred to a clinic for autistic children. As if in a dream, I had found the ideal persons to study. They had been diagnosed with autism and were hypothesized to suffer from a unique form of (men­ tal) illness that needed discovery of a special form of treatment in order to help them. If such treat­ ment was to be discovered and correctly applied, all their problems would quickly vanish. Fortu­ nately, by now I was beginning to learn not to conceptualize clinical problems in that manner. In­ stead, the fascinating part to me was to observe persons with eyes and ears, teeth and toenails, walk­ ing around yet presenting few of the behaviors that one would call social or human. Now, I had the chance to build language and other social and in­ tellectual behaviors where none had existed, a good test of how much help a learning-based approach could offer.

SELF-INJURIOUS BEHAVIORS AND MORE SERENDIPITY

We wasted no time in breaking down the com­ plex construct of autism and developed objective and sensitive measurement systems of the various behaviors ofchildren so diagnosed (Lovaas, Freitag, Gold, & Kassorla, 1965b). In inductive work, the most important first step for any investigator is to secure accurate and sensitive measures of the de-

pendent variables, that is, the client’s various be­ haviors. Sensitive measures will help to assess the impact, ifany, ofthe various independent treatment variables one presents to one’s client. The diagnosis provides only a rough, yes/no assessment. In a field like psychology, where so little is known about what variables affect behavior, sensitive and on-line as­ sessments are critical in order to detect (that is, discover) those environmental variables that may be functional in altering behavior.

In working with children who cannot talk, one soon discovers that they have other problems as well, and one of the most salient of their behaviors is self-injury. Although we had initially intended to establish language, we were forced to search for those variables which controlled self-injury. Oth­ erwise, our study on language could not proceed. Again, we were fortunate because the first client we selected, Beth, had an abundant amount ofself­ injury in addition to extremely limited language. Attempting to help Beth overcome her self-inju­ rious behaviors turned out to be richly instructive. The second break came when we were referred only 1 client during our 1st year. To fill up our labo­ ratory space, we had to work with her from 9 a.m. to 3 p.m. daily, 5 days a week. This gave us ample time to closely observe her and make discoveries. Because we had only 1 subject in our study, we were forced to use a single-subject replication de­ sign. This also turned out to be a major break. Group designs, considered to be the appropriate and acceptable design at that time, would not have helped us isolate the critical variables. Finally, self­ injury is an ideal dependent variable because it can be recorded accurately and laid out over time, be­ coming sensitive to manipulation of treatment vari­ ables.

The study on Beth has been reported elsewhere (Lovaas, Freitag, Gold, & Kassorla, 1965a). I would like to summarize the main lessons we learned, because they came quite unexpectedly and had a profound effect on our subsequent treatment-re­ search project. Beth’s self-injurious behaviors had lasted for about 10 of her 13 years and left her with major scar tissue on her scalp (from banging)

621 TREATMENT-RESEARCH PROJECT

and hands and face (from biting and scratching). She had been institutionalized and received state­ ·of-the-art psychodynamic treatment for at least 1 year, without any apparent change. We discovered that the treatment of choice at that time, derived from psychodynamic theory and supported by com­ mon sense (but no data), centered on delivering sympathetic comments and demonstrations of af­ fection when Beth engaged in self-injurious behav­ iors. Contrary to predictions from this theoretical position, this treatment did in fact accelerate her self-injury. On the other hand, withholding that treatment and building alternate behavior lowered the rate. As we continued to probe with other interventions, we found that major physical and social changes in her environment (such as non­ contingent demonstrations of affection) seemed to have no effect on her. Then, by accident, we dis­ covered that an apparently insignificant change ( changing nursery school songs to those not asso­ ciated with extinction) caused her self-injury to sud­ denly drop to zero, an excellent example ofstimulus control.

The amount of reinforcement we received from Beth during the experimental manipulations was massive, and we came to appreciate and need her more than ever. Her “psychotic episodes” turned out to be rational and social behavior, controlled by known laws that regulated “normal” behaviors. I had up to that point been a “doubting Thomas” when it came to the explanatory power behind reinforcement theory. Increasingly, we came to view the behaviors of developmentally disabled and au­ tistic individuals, not as instances of pathology, but as belonging to the natural order of things (Lovaas & Smith, 1989). Icard (1932; see also Lane, 1977) had worked from this position some 160 years earlier, unencumbered by the many theoretical de­ tours that were to follow him. The research ofWolf, Risley, and Meese ( 1964) and Hewett ( 196 5) par­ alleled our own and was soon to be followed by similar discoveries from other investigators across the country.

One more accidental observation may be worth mentioning, and that concerns the “decision” to

use contingent aversives. Persons often ask me how that came about. When one sees a client (or “sub­ ject”) once or twice a week, one develops a rela­ tionship with that person that is very different from seeing a person 6 hours a day, 5 days a week over most of a year. Also, by the time I saw Beth, I had helped raise four children and learned a great deal about how to raise them. By now I was spend­ ing much more time with Beth than I had with my own children, and I had come to consider her as one of my own. One day, while I briefly inter­ rupted Beth and her teacher’s play to make a short comment, Beth walked away from us to a steel cabinet, bent over, and violently banged her head against the sharp comer. I would not let any of my own children act like that. Quite impulsively and without any contemplation, I reached over and gave her a whack on her behind with my hand. She stopped suddenly and looked at me, as if to ask, “Is this a psychiatric clinic or isn’t it?” I experienced intense fear and guilt as to what I had done. How­ ever, Beth paused for about 1 minute, then as if to test me, hit her head once more. I mustered up enough courage to give her one more slap on the behind. At that point, Beth came back over to the teacher and me, and acted very affectionate and sociable. There were no other acts of self-injury that day, or in my presence thereafter. This incident was never planned, and in fact, I would not have planned to do what I did. It was experiences like these that gradually led us to use the average environment and average children as a model for how to con­ struct a treatment program. These experiences also taught us to change the natural environment only enough to isolate those variables which would make it both therapeutic and educational. As we were learning more about how to help these children, we became increasingly certain that Freud, Bettel­ heim, and others had, by their comprehensive, pop­ ular, and easy-to-understand theories, led them­ selves and others into an enormous blind alley.

The use of contingent aversives has become a controversial issue that warrants more attention than space allows in this paper. Some concerns and pre­ cautions that should be exercised when one con-

622 0. IVAR LOVAAS

templates the use of such interventions have been presented by Lovaas and Favell (1987). Shortly after the work with Beth, we sought out (and were referred) some of the most severely self-injurious clients in the Southern California area. This led us to the large state hospitals where such clients re­ sided. The severity of the problems one can observe in such settings is appalling, and most instructive for anyone who wants to work with developmen­ tally c:lisabled individuals. In short, our observations on the effects of contingent social attention and aversives were replicated (Lovaas & Simmons, 1969). It became increasingly clear that we could severely damage a client by attending to his or her self-injury. The use of contingent aversives quickly suppressed self-injury for most clients, who sub­ sequently could be taken out of restraints and in­ troduced to a large range of educational environ­ ments. I remember vividly, even some 30 years later, a client who learned to walk again after having been tied to his bed for so many years that his tendons had shortened (secondaty to disuse).

Another significant observation occurred when we tried to teach socially appropriate behaviors to these clients. We discovered that most of them, like Beth, were teachable. One client learned the alphabet (receptively) in less than 2 hours! Then, observing the bedlam around them in these large institutions, the thin ratio of staff to clients, and the profound lack of appropriate teaching and treat­ ment skills on part of the staff, one could see it as inevitable that self-injury would develop under such circumstances. The clients had little or no language, and there seemed no other way in which they could control their environment. They all seemed to want some measure of control, much like the rest of us.

One additional illustration of the importance of testing observations from the average environment in designing treatments: In giving one’s child a spank for some destructive behavior, most parents will not leave it at that, but eventually will “make up.” That is, once the child stops the unwanted behavior, the parent waits for (or prompts) some socially appropriate behavior and, when that be­ havior occurs, follows it by assurance to the child

that all was well (a hug and reassurance to the effect that “Ilove you”). In doing so, two important acquisitions may take place. First, socially appro­ priate behavior may be strengthened through neg­ ative reinforcement, because the parent removed all signs of aversives. Second, parental expressions of love may acquire secondaty positive reinforcing properties, becoming ”safety signals” by being as­ sociated with reduction of aversive stimuli. Such use of aversive events is potentially more thera­ peutic than the mere use of their suppressing prop­ erties. Although data supporting these hypotheses were reported long ago (Lovaas, Schaeffer, & Sim­ mons, 1965), these have apparently been ignored or overlooked by those who later implemented aver­ sive interventions.

Given the potential dangers involved in using aversives, we decided to carry out such work with the clinic open to parents, professionals, and the news media. We have maintained an open-door policy since that time, independent of whether we use aversives or not. Such a policy helps to ensure that evetyone works hard, tries to be helpful, and keeps in touch with social ethics. Subsequent in­ volvement of the clients’ parents as active partici­ pants helped to extend that policy and will be discussed later.

Ferster (1961) had postulated that the behav­ ioral deficit of autistic children was due to their deficiency in social reinforcers. Establishing social reinforcers by associating adults with the reduction of aversive stimuli (Lovaas et al., 1965) and/or the presentation of primacy positive reinforcers (such as food) was indeed accomplished (Lovaas et al., 1966). However, this did not by itself result in a concurrent increase in socially appropriate behav­ iors, which had to be separately shaped. Another misleading simplification centered on the possibility of observing major gains in appropriate behaviors with the reduction of self-injurious behaviors. This failed to occur. We were to pursue many other misleading treatment strategies before fully recog­ nizing the complexity of the problem facing us.

Although the causes and treatment of self-in­ jurious behaviors are becoming reasonably well un-

623 TREATMENT-RESEARCH PROJECT

derstood, there are other behavioral excesses among developmentally disabled persons that are not. One such large group of behaviors has been labeled “self-stimulatory,” as observed in the ritualistic, repetitive, stereoryped, and high-rate behaviors of rocking, pacing, jumping up and down, gazing, lining of objects, and so fonh. We have proposed the possibility that such behaviors may be operant behavior, maintained by the sensory-perceptual re­ inforcers that appear to be generated by such be­ haviors (Lovaas, Newsom, & Hickman, 1987). Based on our low success rate in guessing at what the causes of behaviors may be, it is imponant to keep in mind that this is only one of several pos­ sibilities. Some of the treatment implications of these behaviors have been presented elsewhere (Ep­ stein, Taubman, & Lovaas, 1985). These kinds of behaviors are of panicular interest in treatment research, because they appear durable and, unlike most kinds of socially inappropriate behaviors, do not decrease in frequency by withdrawal of social reinforcers. Given the high relapse rate of so many behaviors established by the use of socially medi­ cated reinforcers, it would be a boon to treatment research if one was to discover the variables that created and maintained socially appropriate ”higher levels” of self-stimulatory behaviors. These vari­ ables may not be of the operant kind.

THE 1973 TREATMENT STUDY

The first comprehensive treatment study was be­ gun in 1964 and reflects many of the blind alleys of our own design (Lovaas, Koegel, Simmons, & Long, 1973). First, we worked under the belief that if we removed the children from their natural environment and placed them in an institutional setting, we would be able to obtain very accurate measures of the children’s behaviors on a 24-hour schedule and better control all relevant aspects of their environment. Second, we thought that 1 year of intensive one-to-one treatment (2,000+ hours) would be enough and that treatment gains would last. Finally, we focused our major efforts on de­ veloping language, because we still considered Ian-

guage to be pivotal in facilitating improvement in other nontreated behaviors.

Numerous gains in treatment were made during these years. We observed major increases in com­ plex