The Continuity Model: Alternatives to the Diagnoses

The Continuity Model: Alternatives to the Diagnoses

Traditional theories about persons with autism hypothesize that such persons have something in common that is unique to them, distinguishing them from other groups of persons. This view raises several problems. First, although persons with autism may initially appear to be a rather homogeneous group, closer inspection reveals a wide range of individual differences, casting doubt as to whether these persons really have much, if anything, in common. For example, prior to treatment, individuals with autism may range from showing some mastery of complex language and IQs within the normal range to being mute and scoring within the profoundly retarded range of intellectual functioning (American Psychiatric Association, 1987). They may also vary markedly in their response to treatment. For example, some persons with autism enter treatment with the ability to imitate the speech of others, some acquire speech very rapidly once treatment is begun, some acquire it very slowly, and a few fail to acquire verbal imitation and auditory communication even after extensive training. This final group must be taught visual forms of communication, such as reading and writing or the Picture Exchange Communication System (Chapters 29 and 30, respectively). Similar variability can be observed after treatment. Lovaas (1987) and McEachin, Smith, and Lovaas (1993) reported that three distinct outcome groups emerged when intensive behavioural treatment was given to preschool-aged autistic children: a group that reached normal functioning, an intermediate group that made some gains, and a small residual group that benefited little from the treatment.

The hypothesis that individuals with autism have unique and distinct problems may also be questioned because studies thus far have found that all of the behaviours shown by autistic individuals are also shown by other groups of persons, including normal infants (Rutter, 1978). For example, self-stimulatory behaviours such as rocking and hand flapping, which autistic persons often exhibit at high rates, are quite common in infants (Kravitz & Boehm, 1971). Echolalia, which was once considered a symptom of a psychotic disorder, can be observed in transient form in typical children. Typical children tantrum, and some bang their heads against hard surfaces just like children with autism, albeit to a lesser degree and for shorter periods of time. Indeed, if autistic individuals are matched by mental age to other individuals and their behaviours are compared, most differences disappear (DeMeyer, Hingtgen, & Jackson, 1981).

DeMeyer et al. (1981) and Rutter (1978) wrote excellent reviews of the difficulty posed by individual differences and behavioural overlap. In their reviews, they suggested that the diagnosis of autism may represent a multiplicity of behaviour problems with a multiplicity of aetiologies. Consequently, it is not surprising that efforts to identify causes or effective treatments with traditional approaches have thus far been largely unsuccessful. In essence, the problem is that the existence of an entity called autism is a hypothesis (Rutter, 1978). The tentative-ness of this hypothesis is often overlooked. For example, stating that Leo Kanner was 'the discoverer of autism' (e.g., Schopler, 1987) gives the misleading impression that autism is known to exist. It should be remembered that, like any other hypothesis, autism is a construct that may facilitate research or prematurely freeze or misdirect inquiry into the area of helping the persons to whom the term is applied (Lovaas, 1971b).

In an attempt to understand autistic individuals without pursuing the construct of autism, behaviourists have made three methodological decisions to increase the strength of their research designs and approach to treatment. In short, the large problem, autism, is broken down into smaller units, namely the separate behaviours shown

The majority of this text has been presented in 'A Comprehensive Behavioural Theory of Autistic Children: Paradigm for Research and Treatment,' by O. L Lovaas and T. Smith, 1989, Behaviour Therapy and Experimental Psychiatry, 20, pp. 17-29. by autistic persons, which are reliably and precisely measured. This focus not only allows for precise measurement, it circumvents the problem of autistic persons' behavioural heterogeneity. Because behaviour, rather than autism, is under investigation, behaviour may be studied even if not all autistic persons exhibit it, if different persons exhibit it to different degrees, or if persons without autism sometimes also exhibit it. In fact, such behavioural commonality may facilitate research and treatment of persons with autism because it makes it possible for these persons to be helped by findings from other groups of persons.

Finally, the possibility is left open that each behavioural deviation may have its own neurobiological aetiology, and the delay in the development of a complex behaviour such as language may be the product of many and diverse causes. This implies that remedying different language deviations may require separate and distinct interventions. For example, one kind of intervention may be implemented for a person who does not speak and another for a person with echolalia, and different interventions may be given to individuals who are at the level of acquiring grammar (syntax) in contrast to meaning (semantics).

Psychology and special education may help those with atypical nervous systems by creating and constructing teaching environments by which those with anomalous systems can learn. These special environments should differ as little as possible from the average (everyday) environment for several reasons. First, a primary goal of intervention designed for persons with developmental delays is to help them function more adequately in the typical, everyday environment. The smaller the difference between the special education environment and the average environment, the easier the transfer of skills gained from the special environment to the average environment. Second, the average environment has been developed and 'shaped up' over centuries and contains much information despite its many limitations. Third, by treating individuals with developmental delays as different in degree rather than kind, one may be able to draw upon already acquired knowledge about how typical organisms learn and develop. Such knowledge includes laws of how people learn; a subject about which a great deal of scientific understanding has been acquired. Finally, by breaking down the complex category of autism into its behavioural components, professionals with experience in developing various behaviours, such as language, may be brought in to help with treatment. One need not necessarily be an expert on autism because that hypothetical entity is not addressed. Such was the strategy pursued by investigators who helped develop the treatment programs presented in this teaching manual.

Using the average environment as ^i model, note that a typical (average) individual learns from the beginning of life, all waking hours, every day, for the life of that individual. To approximate this, an appropriate educational environment for a person with developmental delays should be started as early as possible; be in effect most hours of the day, weekends and holidays included; and last for the lifetime of the person involved if so needed. This is in contrast to the current treatment-educational model which intervenes 1 or 2 hours per week (as in speech therapy, psychotherapy, and Sensory Integration) or 6 hours per day, 5 days per week (as in special education). The difference between the intensity of the appropriate intervention and that of the current one is likely to be based on differences in the assumptions underlying these two models of service delivery. If one assumes that there is a central deficit that can be corrected, then a limited intervention that focuses on this deficit may alleviate the problem and set the person free to learn. No empirical evidence exists as of yet, however, to support the beneficial effects of these short-term interventions (Smith, 1993).

Again using the average environment as a model, it should be clear that the intervention should be initiated in the home and involve the individual's parents, relatives, and neighbours. One would want to help the individual acquire a wide range of behaviours (e.g., language and appropriate play) and learning strategies (e.g., imitation) before placing him into a group learning situation such as preschool or kindergarten. Few 2-year-old typical children (with a mental age of 2 years) would be expected to cope with or learn much in a preschool or kindergarten setting.

The behavioural theory proposed in this manual has four tenets, which are outlined here and then described in more detail. First, the laws of learning help account for autistic persons' behaviours and provide the basis for treatment. Second, autistic persons have many separate behavioural deficits best described as developmental delays rather than a central deficit or disease which, if corrected, would lead to broad-based improvement. Consequently, teachers must give them knowledge piece by piece rather than focus on only one broad deficit at a time. Third, autistic persons give evidence of being able to learn like typical individuals if they are placed in special environments. Fourth, autistic persons' failure in normal environments and success in special environments indicate that their problems can be viewed as a mismatch between their nervous system and the environment. At present, given the limited knowledge available from neurobiological research, these persons' problems may be best solved by constructing functional treatment environments.

The Continuity Model

Tenet 1.

Numerous findings indicate that the behaviours of autistic individuals can be accounted for by the laws of learning. When behaviours are reinforced, these individuals show acquisition curves similar to those obtained by other people and other organisms; when reinforcement is withdrawn, the behaviours show extinction curves similar to those gleaned from the behaviours of other organisms (Lovaas, Freitag, Gold, & Kassorla, 1965b). Behaviours that are not acquired in treatment are also related to identifiable reinforcement. For example, self-stimulatory behaviours such as rocking and hand flapping are maintained by the sensory feedback they provide for a person; if this feedback is removed, the behaviours extinguish (Rincover, Newsom, & Carr, 1979). Likewise, self-injurious behaviours and aggression against others (Carr &. Du-rand, 1985) have been found to have one of three functions: They may be self-stimulatory, negatively reinforcing (allowing the individual to escape from aversive situations), or positively reinforcing (leading to attention from others). Prior to treatment, autistic individuals respond to a narrow range of reinforcers, but this range can be expanded through the use of principles derived from learning theory by pairing a stimulus that is neutral for autistic individuals (e.g., praise from others) with another stimulus (e.g., food) that is already reinforcing (Lovaas, Freitag, et al., 1966). Finally, in accordance with behavioural theory, discrimination training paradigms derived from learning theory have been found to be particularly useful in developing treatment programs for autistic individuals (Stoddard & Mcllvane, 1986). Two kinds of discrimination learning provide the basis for teaching much behaviour: imitation and match-to-sample.

Tenet 2.

Autistic individuals have many separate behavioural deficits rather than a central deficit that, if corrected, would lead to broad based change. This tenet is derived from findings on limited response generalization and limited stimulus generalization, as well as the observation that an individual's various behaviours are controlled by different environmental variables.

Most traditional theories of child development hypothesize the existence of an 'organizing construct' (e.g., a 'self,' a 'capacity,' or a 'cognitive schema'), which emerges if a child reaches a certain maturational stage or experiences a particular event or situation. The emergence of this organizing construct creates changes in a wide range of behaviours, leading to what in behavioural terms is referred to as response generalization (changes in behaviours other than those that were specifically taught). In the 1960s, behavioural investigators devoted much effort to finding a pivotal aspect of behaviour that would lead to response generalization in children with autism. However, these efforts were remarkably unsuccessful (Lindsay & Stoffelmayr, 1982). Rather than response generalization, a great deal of response specificity became evident. Changes in such behaviours as language did not result in obvious changes in other behaviours. Even within a seemingly basic or circumscribed unit of behaviour such as the acquisition of abstract language terms, a striking amount of response specificity was shown to be evident. For example, mastery of one class of abstract terms (e.g., prepositions) did not necessarily facilitate understanding of other abstract terms (e.g., pronouns). Children were taught their names or the names of others, but this did not lead them to make a proclamation such as 'different people have different names.' Children were taught to make eye contact and to give and receive affection but, even with these skills, the children remained socially isolated in many ways (e.g., they did not begin to play with others unless explicitly taught to do so).

Like limits in response generalization, limits in stimulus generalization (Stokes & Baer, 1977) provide evidence against the presence of an internal, synthesizing, or organizing capacity. Individuals with autism do not show a capacity 'to take their experiences with them' across environments unless they are directly taught to do so. To remedy this, the individuals must be taught to generalize. For example, Lovaas, Koegel, Simmons, and Long (1973) found that gains made in treatment during hospitalization did not transfer out of the hospital setting unless the children went home and their parents were trained to carry out the behavioural intervention. In more recent studies (Lovaas, 1987; McEachin et al., 1993), many autistic children were able to attain normal levels of functioning at home and school, giving every evidence of generalizing across behaviours and situations. In all likelihood, this came about because the children were taught to acquire information not only from aides working through a clinic, but also from parents at home as well as teachers and peers at school.

Many professionals have objected to behavioural treatment because of issues such as limited response and stimulus generalization, failing to note two important points. First, both kinds of generalization can be taught. Second, the failure to teach rapid generalization may not be characteristic only of behavioural treatment and the failure to obtain such generalization may not be characteristic only of autistic persons. What is the survival value of rapid generalization considering that humankind may experience more failures than successes? A step in the wrong direction may be devastating if it quickly generalizes across other behaviours and environments.

The approach of a behaviour-by-behaviour treatment is the essence of the behavioural approach to the treatment and education of persons diagnosed with autism or mental retardation. This approach leaves open the possibility that the various behavioural delays and excesses associated with autism and mental retardation may be caused by several different kinds of underlying neurological deviations. The treatment, therefore, is not identical across all the individual's behaviours but rather addresses the idiosyncrasies of each behaviour and the uniqueness of each individual.

Behavioural research has proceeded according to an inductive paradigm, and knowledge about treatment has been accumulated in a gradual and systematic manner. Progress in understanding persons diagnosed as having autism or pervasive developmental disorder (PDD) is considered to occur in small steps and gradually rather than suddenly as a result of the discovery of one pivotal problem or disease that controls all behaviours of all persons diagnosed. With behavioural treatment, many persons diagnosed as having autism or PDD achieve normal educational, emotional, social, and intellectual functioning if treatment is started early and given intensively. However, behavioural treatment does not provide a cure for such persons because a cure would have to rectify the root

cause of the problem, which is likely to be various kinds of neurological deviations.

Another finding that counters indicates the presence of a central organizing and generalizing mechanism is that different behaviours have different kinds of lawful relationships with the environment, and even the same behaviour may have different lawful relationships. For example, aggression is sometimes self-stimulatory, sometimes based on negative reinforcement, and sometimes based on positive reinforcement. It seems difficult to invoke some researchable central deficit that could be responsible for such heterogeneity.

Rather than a central deficit, persons with autism appear to have a number of separate behavioural difficulties. These difficulties are best described as developmental delays because, as noted by Rutter (1978), the behaviours are also evident in younger, normal children. Because autistic individuals have so many difficulties that need to be addressed separately, they need to be taught virtually everything, and the teaching needs to proceed in minute increments instead of major steps. Thus, at the beginning of treatment, the individuals may be regarded as being close to a tabula rasa. In this sense, they can be considered very young or recently born, as persons with little or no experience.

Tenet 3.

Persons with autism can ham once a special environment  is constructed for them (Simeonnson, Olley, & Rosenthal, 1987). This special environment should deviate from the normal environment only enough to make it functional. It should consist of regular community settings (as opposed to hospitals, clinics, etc.) and provide requests and consequences for behaviours just as the typical environment does, except that the requests and consequences should be more explicit and 'meaningful' through the use of the learning theory principles discussed in Tenet 1. Preschool-aged children with autism appear to make substantial progress in such an environment (Simeonnson et al., 1987). For example, Lovaas (1987) provided an intensive behavioural intervention (approximately 40 hours of one-on-one treatment per week for many years) to young children with autism. It was found that almost half of the autistic children in Lovaas's sample attained normal levels of intellectual functioning (as measured by IQ tests) and performed satisfactorily in regular first-grade classes by the age of 7 years. These gains were shown to be maintained over time (McEachin et al., 1993).

Tenet 4.

Autistic persons' failure to learn in normal environ' merits and success in special environments indicate that their problems can be viewed as a mismatch between their nervous systems and the normal environment rather than approached as a disease. Because of the dramatic nature of autistic individuals' problems, however, many investigators have been inclined toward dramatic explanations of the problems. From the 1940s to the early 1960s, the problems were attributed to the individuals' parents, who were regarded as extremely hostile. Later on the problems have been attributed to an incurable, organic disorder (DeMeyer et al., 1981). Viewing the individuals' problems as a mismatch between an atypical (though not necessarily diseased) nervous system and an average, day-to-day environment is less dramatic but is more consistent with the data (Lovaas, 1988).

Contrasting Behavioural and Traditional Theories

To clarify the behavioural position further, it may be useful to contrast it with traditional theories of autism. In traditional theories, some structure that is responsible for autistic persons' behavioural deviations is hypothesized to be present. It is this entity that is said to be sick, disordered, or diseased. The aim of treatment is to get inside the person and treat the diseased entity (the 'autism'). If this is done, it is thought that the afflicted persons will begin to live life like others and develop normally. For medically oriented theoreticians, this entity is a neurobiological structure or process, and treatment involves pharmacotherapy, surgery, or other medical interventions. To psychodynamic clinicians, this entity is a self or an ego that must be normalized by refraining from placing demands on the sick individual, and by accepting her and relating through play and fantasy so that a small 'crack in the autistic door' may be opened. This crack would then allow the self to emerge and enable others, such as parents and teachers, to relate to this healthy part of the individual so as to promote growth (Bettelheim, 1967).

Almost all currently available treatment approaches assume the existence of some such internal structure. For example, Holding Therapy (Tinbergen & Tinbergen, 1983) is based on the view that bonding has failed to occur between the mother and the autistic child. Treatment consists of having the mother forcibly hold the child to convey the message that she is available, to alleviate the child's rage and terror, and to cause the 'autistic defence to crumble' (Welch, 1987, p. 48). In contrast, a behavioural position suggests that there may be no intra-psychic conflicts or stress to resolve, no opposing forces, no rage, and no terror of abandonment because the individual never knew of another state. It follows that aides and parents should not feel disappointed when they 'fail to reach the individual' because there was no no autistic person to reach.

Sharing the psychodynamic emphasis on addressing internal states, cognitively oriented psychologists or speech therapists work to stimulate some hypothetical neurological structure to produce more language and other 'higher mental processes' and in turn create and direct new behaviours. A psychomotor therapist also focuses on one aspect of a person's behaviour (physical activities) and hypothesizes that neurological or motivational processes can thereby be activated or normalized. Facilitated Communication makes similar assumptions, as do Fast For Word, Sensory Integration, Option Method, Floor-Time, Play Therapy, Deep Pressure Therapy (e.g., use of squeeze machines), Auditory Integration, Music Therapy, Patterning, Drum Therapy, play with dolphins, horseback riding, 'developmental' models of therapy, rotating chairs, and the many other treatments that have been proposed for individuals with autism.

However diverse these treatments and underlying theoretical systems might seem, they are all based on a belief in certain powerful theoretical variables. These variables, even with brief exposures to treatment in artificial environments, should substantially and permanently improve autistic persons' functioning in all settings. This position allows the therapist to involve only a limited number of professionals in treatment, locate treatment in a clinic or hospital away from the person's community, and set aside a limited number of hours for therapeutic contact. The interventions do not require the therapist or teacher to be familiar with the scientific research forming the basis of these interventions because such research does not exist. The practitioners may need only a few days of training to be qualified to deliver such treatments. Considering all of this, if the position could be supported, it would certainly have some practical advantages. However, data supporting the position have yet to be provided, and findings from behavioural research contradict it.

A related difference between behavioural and traditional theories can be expressed in technical terminology, terminology that will be further elaborated upon in this teaching manual. Behavioural treatment centres on reinforcement control, effecting behaviour change by manipulating the consequences of behaviour. The goal of treatment is to teach a large number of adaptive behaviours (cognitive, linguistic, social, etc.) by reinforcing increasingly closer approximations of the target behaviours and increasingly complex discriminations among situations. In contrast, traditional treatment centres on stimulus control, in which the major manipulations consist of changes in the variables preceding behaviour. The showing of love and acceptance, holding, efforts to arrange a situation that will stimulate speech and physical exercises are all examples of attempts at stimulus control. In short, behavioural treatments aim to build behaviour while traditional treatments aim to cue and stimulate behaviours assumed to be present.

Certainly, cueing behaviour is likely to be easier to administer and result in more rapid improvements than building behaviour. From a behavioural standpoint, however, the main problem with cueing behaviours through stimulus control procedures is that doing so does not result in the acquisition of new behaviours. Stimulus control can only change already existing behaviours, and the behaviours targeted in the treatment of autism are largely in deficit or nonexistent in individuals with autism. Indeed, autistic individuals are identified on the basis of having little or no social behaviour, language, or self-help skills. Thus, from a behavioural standpoint, stimulus control procedures are ineffective for most individuals with autism.

To illustrate the problems with stimulus control, consider a situation in which a teacher puts crayons and a colouring book on a table, seats himself or herself and an autistic child next to the table, and says with a friendly smile and eye contact, 'Let's colour.' The teacher intends for these stimuli (the colouring book, crayons, eye contact, and request) to lead to behavioural changes from which he or she can infer emotional and intellectual growth, such as increased interest in the environment and expressions of creativity. In behavioural terminology, the teacher is attempting to signal, instruct, or otherwise communicate with the child through stimulus control procedures. The main basis for this teaching strategy is probably that it often works with average children. However, with autistic children, the strategy is likely to yield one of the following responses: (a) the child may simply remain seated at the table engaged in self-stimulatory behaviour such as hand flapping; (b) the child may comply with the instruction; or (c) the child may knock over the table and try to bite the teacher, temporarily putting a stop to teaching efforts. In the first and probably most common case, stimulus control is absent (i.e., the stimuli are neutral or non-functional). In the second case, some stimulus control has been established, but it is unclear whether growth will occur. In the third case, stimulus control has also been established, but it yields behaviours that are the opposite of what the teacher intends, perhaps because these behaviours have previously been reinforced and shaped by negative reinforcement (termination of teaching sessions).

In this example, the teacher uses an easily understood, commonsense intervention that has the advantage of being supported by experience with other, more average individuals and by various traditional theories of development. However, the results of this intervention are likely to be disconcerting and disappointing, as there are no empirical data from controlled experiments to indicate that autistic individuals benefit from such interventions. By contrast, behavioural intervention requires more technical knowledge concerning stimulus and reinforcement control. Therefore, it is more complicated to understand and implement, but it comes closer to addressing the problems that autistic individuals present.

The parents and professionals who have to make decisions about a person's treatment are advised to raise the following questions: First, have outcome data been published in professional journals with peer review, and what other scientific research forms the basis for the intervention? Second, what kind of training has the provider received in administering the intervention, whether behavioural or not, and how long did this training last? If the answer is a workshop lasting a week or less with occasional consultations, remain highly sceptical of the adequacy of the services the individual will receive because the problems the individual is facing are likely to be more complex than the provider has training to handle.

One may wonder why it is difficult to give up the belief in a controlling and organizing entity that can be fixed and thus allow the individual freedom to develop. One argument for the existence of such organizing and facilitating brain mechanisms emphasizes the ease of application and efficiency of the intervention. With this argument, there is a more subtle and seductive promise involved: We human beings are born with all kinds of innate skills or capacities, such as prewired grammar (a la Chomsky), morality (a la Kohlberg), and maturational readiness for quick assimilation of cognitive skills (a la Piaget), once exposure has been achieved. The promise of these philosophical positions is that teachers and parents alike have to do less for our children; our nature of being human has a way of steering us on the right track. Such a promise may be false. The unfortunate consequence for individuals with autism and other developmental delays and for typical children as well, is that less can be done to help them.

How may a behaviourist offer hope? First, consider the individual with autism to be a very young child. Young children do not give evidence of knowing much about the world around them. Second, consider the individual as having failed to develop in the average environment which has taught others, but capable of developing and growing in a special teaching environment as described in this manual. Whether the growth is small or large, satisfaction may be gained. As a parent once said, 'My child's daily progress, however small, is my reinforcer.'

If all of a person's behaviours become normalized, should the person still be considered autistic or retarded? A behaviourist would likely answer, 'No.' Many other professionals would likely consider such a situation as 'autism in residual state,' reflecting the pervasive hold that concepts such as autism and retardation exercise over treatment and research. A good illustration of this traditional position is provided by a teacher who observed a child she had first encountered shortly after the diagnosis of autism was rendered to that child. Two years of 40 hours per week of intensive behavioural treatment later, the teacher observed the same child in a typical class and exclaimed, 'I looked and looked for the last couple of days and I want to know: Where is the autism?'

Ethical Considerations

Rather than the placement of atypical persons into discrete diagnostic categories such as autistic, schizophrenic, or mentally retarded, such persons can be viewed as differing in degree and as contributing to the variability of humankind. All living systems possess variability, and variability is essential for physical survival as well as for new directions in science and art. A society that restricts variability (e.g., regimes such as those proposed by Marx and Hitler) is at a disadvantage in facing the future because it does not possess the variability and flexibility necessary for coping with new environments that demand new behaviours in a future we cannot predict.

We came to view atypical persons as belonging to this continuum of variability, as different in degree rather than kind. One can accept such variability not only because of its potential contribution to survival in future environments, but also because it allows the rest of us to remain as we are. Those who differ are our protection in an uncertain future. For this we can admire and appreciate them.

Encouraging variability raises questions about whether to treat atypical persons. Consider Van Gogh, whose art influenced and helped define our culture. His lack of social skills is well known and pronounced; today he may have been diagnosed as having autistic features or having schizophrenia. What a sad fate for us all if we had successfully 'treated' him. Successful treatment that centred on helping Van Gogh acquire social skills would have sensitized him to a large range of new rewards. Consequently, he may have left his socially isolated existence and spent time with friends and lovers, leaving little or no time to paint. The same fate may have been met by many other of our outstanding artists and scientists. Einstein, a social isolate, is a case in point. Why, then, treat socially isolated persons with autism? Because without a more varied behavioural repertoire, these persons could not survive on their own. A person with a varied behavioural repertoire has more options and is more likely to survive. In the case of autism, with its severe and multiple behavioural delays, nature may have overshot its mark.

Directions for Future Behavioural Work

Although we believe that the behavioural approach described in this manual addresses many of the questions posed through investigations into autism, we recognize that many other important questions remain unanswered and require further research. Answers to these questions will probably add to, rather than replace, the present approach because this approach is based on inductive, cumulative research instead of a specific hypothesis.

Some of the questions that remain are more practical than theoretical. For example, effective implementation of behavioural procedures requires a major reorganization of the way in which treatment is now delivered: Those who are best suited to deliver such treatment (e.g., special education teachers) need to receive instruction in the specialized skills required to do behavioural work with autistic persons, work in the home and community rather than in clinical settings or schools, change their curriculum, and collaborate closely with all significant individuals (including parents, teachers, siblings, and friends) who interact with the persons treated. Even though these changes might be very difficult to implement, they would result in substantial benefits for persons with autism and their families.

Although the improved functioning of persons with autism is clearly one benefit of intensive behavioural treatment, other practical benefits should not be overlooked.

Providing the treatment may require assigning one professional (e.g., a special education teacher) and several aides to work with one individual on a full-time basis for 2 years, which would currently cost approximately $120,000 and might enable about one half of young autistic children to attain normal levels of functioning. For each person who attains normal functioning, it is likely that more than $2 million in treatment costs alone would be saved by preventing lifelong supervision (special education classes, hospital admissions, residential placements, etc.).

Some problems that remain are both practical and theoretical, such as the development of instruments for identifying children with autism during the first or second year of life. Such identification would help investigators study the early history of autistic children's problems directly instead of having to rely on parents' recollections (cf. Rutter & Lockyer, 1967). Further, early identification may be the key to increasing the proportion of autistic children who attain normal functioning. For example, the group of children who did not reach normal functioning in Lovaas's (1987) Young Autism Project might have succeeded had treatment been started earlier.

Another treatment question is whether the interventions devised for individuals with autism are applicable to persons of other diagnostic groups. For example, persons with schizophrenia have been regarded as quite distinct from persons with autism (see, e.g., Rutter, 1978); however, from a behavioural standpoint, individuals diagnosed as having autistic disorder and individuals who develop schizophrenia, Asperger's syndrome, PDD not otherwise specified, and attention deficit disorder appear to have many similar behaviour problems (e.g., poor social skills, attention deficits, language and cognitive delays, stereotyped behaviours). Because the latter persons are less delayed in development than persons with autism, it seems that they may make progress with behavioural intervention. This would be important to determine because there has been very little progress in the treatment of persons from these other diagnoses.

Research that may offer new conceptual insights can be illustrated in the area of self-stimulatory (ritualistic and high-rate) behaviours. A wide variety of important behavioural phenomena appear to be self-stimulatory. For example, some instances of self-injurious behaviour have been found to be self-stimulatory in nature (Favell et al., 1982). Echolalia speech also possesses characteristics of self-stimulatory behaviour: It is repetitive, high rate, and shows resistance to extinction. Another conceptual insight provided by research into the area of self-stimulatory behaviour is that many persons with autism who undergo intensive behavioural treatment begin to perseverate on spelling certain words, reciting numbers, memorizing calendars, and so on, when they are first exposed to these stimuli (Epstein, Taubman, & Lovaas, 1985). Why certain kinds of materials become part of some persons' self-stimulatory behaviour remains a puzzle, but this phenomenon is important to research because it may facilitate treatment and because it is not well explained by learning theory as of yet.

As presented in our theory, individuals with autism generally develop slowly, and skills have to be taught explicitly. How does this development compare with the development of typical individuals? Most developmental theories such as that of Piaget (Flavell, 1963) posit stages during which children quickly and spontaneously transform in a number of ways as a result of a change in cognitive structure. Such theories describe a very different process from what behavioural studies have found in children with autism, but these theories have had considerable influence on the treatment of children with autism (e.g., Schopler & Reichler, 1976). Therefore, it is of interest to find out whether differences between these theories and behavioural data reflect actual differences in development or merely differences in theoretical orientation.

Another area of research that may lead to conceptual insights is further examination into links between behavioural treatment and neurology. Intensive behavioural intervention may act to reverse the neurological problems involved in autistic children's problems, especially considering that environmental interventions have been shown to bring about major changes in neurological structure under some circumstances, particularly in infants (Neville, 1985). Through investigation into this area, we may come close to providing a 'cure' for the behavioural difficulties demonstrated by children with autism. The interventions applied through these studies may also prevent difficulties that have been found to emerge in autistic persons later in life, such as seizures and high serotonin levels (W. H. Green, 1988). Recent advances in brain imaging techniques (Sokoloff, 1985) may facilitate the study of brain environment relationships in persons with autism. This in turn would open up new research areas and contribute to the understanding of the neurological deviations that underlie autism and the effects of environmental interventions on neurological activity.