History And Background Of Autism Discussion essay
Journal of Consulting and Oimcal Ps~hology Copyright !987 by the American P!ychological Association. Inc. 1987, Vol. 55, No. I, 3-9 0022-006X/87 /SOO. 7 5 Behavioral Treatment and Normal Educational and Intellectual Functioning in Young Autistic Children 0. I var Lovaas University of California, Los Angeles Autism is a serious psychological disorder with onset in early childhood. Autistic children show minimal emotional attachment, absent or abnormal speech, retarded IQ, ritualistic behaviors, ag­ gression, and self-injury. The prognosis is very poor, and medical therapies have not proven effective. This article reports the results of behavior modification treatment for two groups of similarly consti­ tuted, young autistic children. Follow-up data from an intensive, long-term experimental treatment group (n = 19) showed that 47% achieved normal intellectual and educational functioning, with normal-range IQ scores and successful first grade performance in public schools. Another 40% were mildly retarded and assigned to special classes for the language delayed, and only !0% were pro­ foundly retarded and assigned to classes for the autistic/retarded. In contrast, only 2% of the control­ group children (n = 40) achieved normal educational and intellectual functioning; 45% were mildly retarded and placed in language-delayed classes, and 53% were severely retarded and placed in autis­ tic/retarded classes. Kanner ( 1943) defined autistic children as children who ex­ hibit (a) serious failure to develop relationships with other peo­ ple before 30 months of age, (b) problems in development of normal language, (c) ritualistic and obsessional behaviors ("in­ sistence on sameness"), and (d) potential for normal intelli­ gence. A more complete behavioral definition has been pro­ vided elsewhere (Lovaas, Koegel, Simmons, & Long, 1973). The etiology of autism is not known, and the outcome is very poor. In a follow-up study on young autistic children, Rutter ( 1970) reported that only 1.5% of his group (n = 63) had achieved nor­ mal functioning. About 35% showed fair or good adjustment, usually required some degree of supervision, experienced some difficulties with people, had no personal friends, and showed minor oddities of behavior. The majority (more than 60%) re­ mained severely handicapped and were living in hospitals for mentally retarded or psychotic individuals or in other protective settings. Initial IQ scores appeared stable over time. Other stud­ ies (Brown, 1969; DeMyer et al., 1973; Eisenberg, 1956; Free­ man, Ritvo, Needleman, & Yokota, l 985; Havelkova, 1968) re- This study was supported by Grant MH-11440 from the National Institute of Mental Health. Aspects of this study were presented at the 1982 convention of the American Psychological Association, Washing­ ton, DC, by Andrea Ackerman, Paula Firestone, Gayle Goldstein, Ron­ ald Leaf. John McEachin, and the author. The author expresses his deep appreciation to the many undergraduate students at the University of California, Los Angeles, who served as student therapists on the project, to the many graduate students who served as clinic supervisors, and to the many parents who trusted their children to our care. Special thanks to Laura Schreibman and Robert Koegel, who collaborated in the early stages of this research project. Donald Baer, Bruce Baker, Bradley Bucher, Arthur Woodward, and Haikang Shen provided statistical ad­ vice and help in manuscript preparation. B. J. Freeman's help in arrang­ ing access to Control Group 2 data is also appreciated. Correspondence concerning this article should be addressed to 0. Ivar Lovaas, Psychology Department, University of California, 405 Hilgard Avenue, Los Angeles, California 90024. port similar data. Higher scores on IQ tests, communicative speech, and appropriate play are considered to be prognostic of better outcome (Lotter, 1967). Medically and psychodynamically oriented therapies have not proven effective in altering outcome (DeMyer, Hingtgen, & Jackson, 1981 ). No abnormal environmental etiology has been identified within the children's families (Lotter, 1967). At pres­ ent, the most promising treatment for autistic persons is behav­ ior modification as derived from modem learning theory (De­ Myer et al., 1981 ). Empirical results from behavioral interven­ tion with autistic children have been both positive and negative. On the positive side, behavioral treatment can build complex behaviors, such as language, and can help to suppress pathologi­ cal behaviors, such as aggression and self-stimulatory behavior. Clients vary widely in the amount of gains obtained but srow treatment gains in proportion to the time devoted to treatment. On the negative side, treatment gains have been specific to the particular environment in which the client was treated, sub­ stantial relapse has been observed at follow-up, and no client has been reported as recovered (Lovaas et al., 1973). The present article reports a behavioral-intervention project (begun in 1970) that sought to maximize behavioral treatment gains by treating autistic children during most of their waking hours for many years. Treatment included all significant per­ sons in all significant environments. Furthermore, the project focused on very young autistic children (below the age of 4 years) because it was assumed that younger children would be less likely to discriminate between environments and therefore more likely to generalize and to maintain their treatment gains. Finally, it was assumed that it would be easier to successfully mainstream a very young autistic child into preschool than it would be to mainstream an older autistic child into primary school. It may be helpful to hypothesize an outcome of the present study from a developmental or learning point of view. One may assume that normal children learn from their everyday environ- 3 4 0. !VAR LOVAAS ments most of their waking hours. Autistic children, conversely, do not learn from similar environments. We hypothesized that construction of a special, intense, and comprehensive learning environment for very young autistic children would allow some of them to catch up with their normal peers by first grade. Method Subjects Subjects were enrolled for treatment if they met three criteria: (a) independent diagnosis of autism from a medical doctor or a licensed PhD psychologist, (b) chronological age (CA) less than 40 months if mute and less than 46 months if echolalic, and ( c) prorated mental age (PMA) of 11 months or more at a CA of 30 months. The last criterion excluded 15% of the referrals. The clinical diagnosis of autism emphasized emotional detachment, extreme interpersonal isolation, little if any toy or peer play, language disturbance (mutism or echolalia), excessive rituals, and onset in in­ fancy. The diagnosis was based on a structured psychiatric interview with parents, on observations of the child's free-play behaviors, on psy­ chological testing of intelligence, and on access to pediatric examina­ tions. Overthe 15 years of the project, the exact wording of the diagnosis changed slightly in compliance with changes in the Diagnostic and Sta­ tistical Manual of Mental Disorders (DSM-III; American Psychiatric Association, 1980). During the last years, the diagnosis was made in compliance with DSM-Ill criteria (p. 87). In almost all cases, the diag­ nosis of autism had been made prior to family contact with the project. Except for one case each in the experimental group and Control Group I, all cases were diagnosed by staff of the Department of Child Psychia­ try, University of California, Los Angeles (UCLA) School of Medicine. Members of that staff have contributed to the writing of the DSM-III and to the diagnosis of autism adopted by the National Society for Chil­ dren and Adults with Autism. If the diagnosis of autism was not made, the case was referred elsewhere. In other words, the project did not select its cases. More than 90% of the subjects received two or more indepen­ dent diagnoses, and agreement on the diagnosis of autism was 100%. Similarly high agreement was not reached for subjects who scored within the profoundly retarded range on intellectual functioning (PMA < 11 months); these subjects were excluded from the study. Treatment Conditions Subjects were assigned to one of two groups: an intensive-treatment experimental group (n = 19) that received more than 40 hours of one­ to-one treatment per week, or the minimal-treatment Control Group I (n = 19) that received IO hours or less of one-to-one treatment per week. Control Group I was used to gain further information about the rate of spontaneous improvement in very young autistic children, especially those selected by the same agency that provided the diagnostic work-up for the intensive-treatment experimental group. Both treatment groups received treatment for 2 or more years. Strict random assignment (e.g., based on a coin Hip) to these groups could not be used due to parent protest and ethical considerations. Instead, subjects were assigned to the experimental group unless there was an insufficient number of staff members available to render treatment (an assessment made prior to contact with the family). Two subjects were assigned to Control Group I because they lived further away from UCLA than a I-hr drive, which made sufficient staffing unavailable to those clients. Because fluctua­ tions in staff availability were not associated in any way with client char­ acteristics, it was assumed that this assignment would produce unbiased groups. A large number of pretreatment measures were collected to test this assumption. Subjects did not change group assignment. Except for two families who left the experimental gr~up within the first 6 months (this group began with 21 subjects), all families stayed with their groups from beginning to end. Assessments Pretreatment mental age (MA) scores were based on the following scales (in order of the frequency of their use): the Bayley Scales of Infant Development (Bayley, 1955), the Cattell Infant Intelligence Scale (Cat­ tell, 1960), the Stanford-Binet Intelligence Scale(Thorndike, 1972), and the Gesell Infant Development Scale (Gesell, 1949). The first three scales were administered to 90% of the subjects, and relative usage of these scales was similar in each group. Testing was carried out by gradu­ ate students in psychology who worked under the supervision of clinical psychologists at UCLA or licensed PhD psychologists at other agencies. The examiner chose the test that would best accommodate each sub­ ject's developmental level, and this decision was reached independently of the project staff. Five subjects were judged to be untestable (3 in the experimental group and 2 in Control Group I). Instead, the Vineland Social Maturity Scale (Doll, 19 5 3) was used to estimate their MAs ( with the mother as informant). To adjust for variations in MA scores as a function of the subject's CA at the time of test administration, PMA scores were calculated for a CA at 30 months (MA/CA X 30). Behavioral observations were based on videotaped recordings of the subject's free-play behavior in a playroom equipped with several simple early-childhood toys. These videotaped recordings were subsequently scored for amount of (a) selfstimulatory behaviors. defined as pro­ longed ritualistic, repetitive, and stereotyped behavior such as body­ rocking, prolonged gazing at lights, excessive hand-flapping, twirling the body as a top, spinning or lining of objects, and licking or smelling of objects or wall surfaces; (b) appropriate play behaviors, defined as those limiting the use of toys in the playroom to their intended purposes, such as pushing the truck on the floor, pushing buttons on the toy cash regis­ ter, putting a record on the record player, and banging with the toy ham­ mer; and (c) recognizable words. defined to include any recognizable word, independent of whether the subject used it in a meaningful con­ text or for communicative purposes. One observer who was naive about subjects' group placement scored all tapes after being trained to agree with two experienced observers (using different training tapes from sim­ ilar subjects). Interobserver reliability was scored on 20% of the tapes (randomly selected) and was computed for each category of behavior for each subject by dividing the sum of observer agreements by the sum of agreements and disagreements. These scores were then summed and averaged across subjects. The mean agreement (based both on occur­ rences and nonoccurrences) was 91 % for self-stimulatory behavior, 85% for appropriate play behavior, and I 00% for recognizable words. A more detailed description of these behavioral recordings has been provided elsewhere(Lovaasetal., 1973). A I-hr parent interview about the subjects' earlier history provided some diagnostic and descriptive information. Subjects received a score of I for each of the following variables parents reported: no recognizable words; no toy play (failed to use toys fortheirintended function); lack of emotional attachment (failed to respond to parents' affection); apparent sensory deficit (parents had suspected their child to be blind or deaf because the child exhibited no or minimal eye contact and showed an unusually high pain threshold); no peer play (subject did not show inter­ active play with peers); self-stimulatory behavior; tantrums (aggression toward family members or self); and no toilet training. These 8 mea­ sures from parents' intake interviews were summed to provide a sum pathology score. The intake interview also provided information about abnormal speech (0 = normal and meaningful language, however lim­ ited; I = echolalic language used meaningfully [e.g., to express needs]; 2 = echolalia; and 3 ~ mute); age of walking; number of siblings in the family; socioeconomic status of the father; sex; and neurological examinations (including EEGs and CAT scans) that resulted in findings of pathology. Finally, CA at first diagnosis and at the beginning of the 5 TREATMENT OF AUTISTIC CHILDREN present treatment were recorded. This yielded a total of20 pretreatment measures, 8 of which were collapsed into I measure (sum pathology). A brief clinical description of the experimental group at intake fol­ lows (identical to that for Control Group I): Only 2 of the 19 subjects obtained scores within the normal range of intellectual functioning; 7 scored in the moderately retarded range, and IO scored in the severely retarded range. No subject evidenced pretend or imaginary play, only 2 evidenced complex (several different or heterogeneous behaviors that together formed one activity) play, and the remaining subjects showed simple (the same elementary but appropriate response made repeat­ edly) play. One subject showed minimal appropriate speech, 7 were echolalic, and 11 were mute. According to the literature that describes the developmental delays of autistic children in general, the autistic sub­ jects in the present study constituted an average (or below average) sam­ ple of such children. Posttreatment measures were recorded as follows: Between the ages of6 and 7 years (when a subject would ordinarily have completed first grade), information about the subjects' first-grade placement was sought and validated; about the same time, an IQ score was obtained. Testing was carried out by examiners who were naive about the subjects' group placement. Different scales were administered to accommodate differ­ ent developmental levels. For example, a subject with a regular educa­ tional placement received a Wechsler Intelligence Scale for Children­ Revised (WISC-R; Wechsler, 1974) or a Stanford-Binet Intelligence Scale (Thorndike, 1972), whereas a subject in an autistic/retarded class received a nonverbal test like the Merrill-Palmer Pre-School Perfor­ mance Test (Stutsman, 1948). In all instances of subjects having achieved a normal IQ score, the testing was eventually replicated by other examiners. The scales (in order of the frequency of usage) in­ cluded the WISC-R (Wechsler, 1974), the Stanford-Binet (Thorndike, 1972), the Peabody Picture Vocabulary Test (Dunn, 1981 ), the Wech­ sler Pre-School Scale (Wechsler, 1967), the Bayley Scales oflnfant De­ velopment (Bayley, 1955), the Cattell Infant Intelligence Scale (Cattell, 1960), and the Leiter International Performance Scale (Leiter, 1959). Subjects received a score of 3 for normal functioning if they received a score on the WISC-R or Stanford-Binet in the normal range, completed first grade in a normal class in a school for normal children, and were advanced to the second grade by the teacher. Subjects received a score of 2 if they were placed in first-grade in a smaller aphasia (language delayed, language handicapped, or learning disabled) class. Placement in the aphasia class implied a higher level offunctioning than placement in classes for the autistic/retarded, but the diagnosis of autism was al­ most always retained. A score of I was given if the first-grade placement was in a class for the autistic/retarded and if the child's IQ score fell within the severely retarded range. Treatment Procedure Each subject in the experimental group was assigned several well trained student therapists who worked (part-time) with the subject in the subject's home, school, and community for an average of 40 hr per week for 2 or more years. The parents worked as part of the treatment team throughout the intervention; they were extensively trained in the treatment procedures so that treatment could take place for almost all of the subjects' waking hours, 365 days a year. A detailed presentation of the treatment procedure has been presented in a teaching manual (Lovaas et al., 1980). The conceptual basis of the treatment was rein­ forcement (operant) theory; treatment relied heavily on discrimination­ learning data and methods. Various behavioral deficiencies were tar­ geted, and separate programs were designed to accelerate development for each behavior. High rates of aggressive and self-stimulatory behav­ iors were reduced by being ignored; by the use of time-out; by the shap­ ing of alternate, more socially acceptable forms of behavior; and (as a last resort) by the delivery of a loud "no" or a slap on the thigh contin­ gent upon the presence of the undesirable behavior. Contingent physical aversives were not used in the control group because inadequate staffing in that group did not allow for adequate teaching of alternate, socially appropriate behaviors. During the first year, treatment goals consisted of reducing self-stimu­ latory and aggressive behaviors, building compliance to elementary ver­ bal requests, teaching imitation, establishing the beginnings of appro­ priate toy play, and promoting the extension of the treatment into the family. The second year of treatment emphasized teaching expressive and early abstract language and interactive play with peers. Treatment was also extended into the community to teach children to function within a preschool group. The third year emphasized the teaching of appropriate and varied expression of emotions; preacademic tasks like reading, writing, and arithmetic; and observational learning (learning by observing other children learn). Subjects were enrolled only in those preschools where the teacher helped to carry out the treatment pro­ gram. Considerable effort was exercised to mainstream subjects in a normal (average and public) preschool placement and to avoid initial placement in special education classes with the detrimental effects of exposure to other autistic children. This occasionally entailed withhold­ ing the subject's diagnosis ofautism. If the child became known as autis­ tic (or as "a very difficult child") during the first year in preschool, the child was encouraged to enroll in another, unfamiliar school (to start fresh). After preschool, placement in public education classes was deter­ mined by school personnel. All children who successfully completed normal kindergarten successfully completed first grade and subsequent normal grades. Children who were observed to be experiencing educa­ tional and psychological problems received their school placement through Individualized Educational Plan (IEP) staffings (attended by educators and psychologists) in accordance with the Education For All Handicapped Children Act of 197 5. All subjects who went on to a normal first grade were reduced in treatment from the 40 hr per week characteristic of the first 2 years to IO hr or less per week during kindergarten. After a subject had started first grade, the project maintained a minimal (at most) consultant rela­ tionship with some families. In two cases, this consultation and the sub­ sequent correction of problem behaviors were judged to be essential in maintaining treatment gains. Subjects who did not recover in the experimental group received 40 hr or more per week of one-to-one treat­ ment for more than 6 years (more than 14,000 hr of one-to-one treat­ ment), with some improvement shown each year but with only I subject recovering. Subjects in Control Group I received the same kind of treatment as those in the experimental group but with less intensity (less than I 0 hr of one-to-one treatment per week) and without systematic physical aversives. In addition, these subjects received a variety of treatments from other sources in the community such as those provided by small special education classes. Control Group 2 consisted of 21 subjects selected from a larger group (N = 62) of young autistic children studied by Freeman et al. ( 1985). These subjects came from the same agency that diagnosed 95% of our other subjects. Data from Control Group 2 helped to guard against the possibility that subjects who had been referred to us for treatment con­ stituted a subgroup with particularly favorable or unfavorable out­ comes. To provide a group of subjects similar to those in the experimen­ tal group and Control Group 1, subjects for Control Group 2 were se­ lected if they were 42 months old or younger when first tested, had IQ scores above 40 at intake, and had follow-up testing at 6 years of age. These criteria resulted in the selection of 21 subjects. Subjects in Con­ trol Group 2 were treated like Control Group 1 subjects but were not treated by the Young Autism Project described here. Results Pretreatment Comparisons Eight pretreatment variables from the experimental group and Control Group I (CA at first diagnosis, CA at onset of treat- 6 0. IVAR LOVAAS Table I Means and F Ratios From Comparisons Between Groups on Intake Variables Recognizable Toy Self- Sum Abnormal Group Diagnosis CA Treatment CA PMA words play stimulation pathology speech Experimental 32.0 34.6 18.8 .42 28.2 12.1 6.9 2.4 Control I 35.3 40.9 17.1 .58 20.2 19.6 6.4 2.2 F' 1.58 4.02* 1.49 .92 2.76 3.37 .82 .36 Note. CA = chronological age; PMA = prorated mental age. Experimental group, n = 19; Control Group I, n = 19. 'df= l, 36. • p < .05. ment, PMA, sum pathology, abnormal speech, self-stimulatory behavior, appropriate toy play, and recognizable words) were subjected to a multivariate analysis of variance (MANOVA; Brecht & Woodward, 1984 ). The means and F ratios from this analysis are presented in Table I. As can be seen, there were no significant differences between the groups except for CA at on­ set of our treatment (p < .05). Control subjects were 6 months older on the average than experimental subjects (mean CAs of 35 months vs. 41 months, respectively). These differences prob­ ably reflect the delay of control subjects in their initiation into the treatment project because of staff shortages; analysis will show that differential CAs are not significantly related to out­ come. To ascertain whether another test would reveal a statisti­ cally significant difference between the groups on toy play, de­ scriptions of the subjects' toy play (taken from the videotaped recordings) were typed on cards and rated for their developmen­ tal level by psychology students who were naive about the pur­ pose of the ratings and subject group assignment. The ratings were reliable among students (r = .79, p < .001), and an Ftest showed no significant difference in developmental levels of toy play between the two groups. The respective means from the experimental group and Con­ trol Group I on the eight variables from the parent interview were .89 and .74 for sensory deficit, .63 and .42 for adult rejec­ tion, .58 and .47 for no recognizable words, .53 and .63 for no toy play, 1.0 and 1.0 for no peer play, .95 and .89 for body self­ stimulation, .89 and .79 for tantrums, and .68 and .63 for no toilet training. The experimental group and Control Group I were also similar in onset of walking (6 vs. 8 early walkers; I vs. 2 late walkers), number of siblings in the family ( 1.26 in each group), socioeconomic status of the father (Level 49 vs. Level 54 according to 1950 Bureau of the Census standards), boys to girls ( 16:3 vs. 11 :8); and number of subjects referred for neuro­ logical examinations (IO vs. 15) who showed signs of damage (0 vs. I). The numbers of favorable versus unfavorable prognostic signs (directions of differences) on the pretreatment variables divide themselves equally between the groups. In short, the two groups appear to have been comparable at intake. Follow-Up Data Subjects' PMA at intake, follow-up educational placement, and IQ scores were subjected to a MANOV A that contrasted the experimental group with Control Groups I and 2. At intake, there were no significant differences between the experimental group and the control groups. At follow-up, the experimental group was significantly higher than the control groups on educa- tional placement (p < .001) and IQ (p < .01). The two control groups did not differ significantly at intake or at follow-up. In short, data from Control Group 2 replicate those from Control Group I and further validate the effectiveness of our experi­ mental treatment program. Data are given in Table 2 that show the group means from pretreatment PMA and posttreatment educational placement and IQ scores. The table also shows the F ratios and significance levels of the three group comparisons. In descriptive terms, the 19-subject experimental group shows 9 children (47%) who successfully passed through nor­ mal first grade in a public school and obtained an average or above average score on IQ tests (M = 107, range = 94-120). Eight subjects ( 42%) passed first grade in aphasia classes and obtained a mean IQ score within the mildly retarded range of intellectual functioning (M = 70, range = 56-95). Only two children ( I 0%) were placed in classes for autistic/retarded chil­ dren and scored in the profoundly retarded range (IQ < 30). There were substantial increases in the subjects' levels of in­ tellectual functioning after treatment. The experimental group subjects gained on the average of 30 IQ points over Control Group I subjects. Thus the number of subjects who scored within the normal range of intellectual functioning increased from 2 to 12, whereas the number of subjects within the moder­ ate-to-severe range of intellectual retardation dropped from I 0 to 3. As of 1986, the achievements of experimental group sub- Table2 Means and F Ratios for Measures at Pretreatment and Posttreatment Follow-up Group Intake PMA EDP IQ Means Experimental Control I Control 2 18.8 17.1 17.6 2.37 1.42 1.57 83.3 52.2 57.5 Fratios' Experimental X Control I Experimental X Control 2 Control I X Control 2 1.47 0.77 0.14 23.6** 17.6** 0.63 14.4·· I0.4* 0.45 Note. PMA = prorated mental age; EDP = educational placement. Ex­ perimental group, n = 19; Control Group I, n = 19; Control Group 2, n = 21. 'df= 1,56. * p < .01. •• p < .001. 7 TREATMENT OF AUTISTIC CHILDREN Table3 Educational Placement and Mean and Range of IQ at Follow- Up Group Recovered Aphasic Autistic/Retarded Experimental N 9 8 2 MlQ Range 107 94-120 70 56-95 30 -· Control Group I N 0 8 II MIQ 74 36 Range 30-102 20-73 Control Group 2 N l 10 10 MlQ 99 67 44 Range 49-8! 35-54 Note. Dashes indicate no score or no entry. ' Both children received the same score. jects have remained stable. Only 2 subjects have been reclassi­ fied: l subject (now 18 years old) was moved from an aphasia to a normal classroom after the sixth grade; I subject (now 13 years old) was moved from an aphasia to an autistic/retarded class placement. The MA and IQ scores of the two control groups remained virtually unchanged between intake and follow-up, consistent with findings from other studies (Freeman et al., 1985; Rutter, 1970). The stability of the IQ scores of the young autistic chil­ dren, as reported in the Freeman et al. study, is particularly relevant for the present study because it reduces the possibility of spontaneous recovery effects. In descriptive terms, the com­ bined follow-up data from the control groups show that tbeir subjects fared poorly: Only I subject (2%) achieved normal functioning as evidenced by normal first-grade placement and an IQ of99 on the WISC-R; 18 subjects (45%) were in aphasia classes(mean IQ= 70, range= 30-101); and 21 subjects(53%) were in classes for the autistic/retarded (mean IQ= 40, range = 20-73). Table 3 provides a convenient descriptive summary of the main follow-up data from the three groups. One final control procedure subjected 4 subjects in the exper­ imental group (Ackerman, 1980) and 4 subjects in Control Group l (McEachin & Leaf, 1984) to a treatment intervention in which one component of treatment (the loud "no" and occa­ sional slap on the thigh contingent on self-stimulatory, aggres­ sive, and noncompliant behavior) was at first withheld and then introduced experimentally. A within-subjects replication de­ sign was used across subjects, situations, and behaviors, with baseline observations varying from 3 weeks to 2 years after treatment had started (using contingent positive reinforcement only). During baseline, when the contingent-aversive compo­ nent was absent, small and unstable reductions were observed in the large ;mount of inappropriate behaviors, and similar small and unstable increases were observed in appropriate be­ haviors such as play and language. These changes were insuffi­ cient to allow for the subjects' successful mainstreaming. Intro­ duction of contingent aversives resulted in a sudden and stable reduction in the inappropriate behaviors and a sudden and sta­ ble increase in appropriate behaviors. This experimental inter­ vention helps to establish two points: First, at least one compo- nent in the treatment program functioned to produce change, which helps to reduce the effect of placebo variables. Second, this treatment component affected both the experimental and control groups in a similar manner, supporting the assumption that the two groups contained similar subjects. Analyses of variance were carried out on the eight pretreat­ ment variables to determine which variables, if any, were sig­ nificantly related to outcome (gauged by educational placement and IQ) in the experimental group and Control Group I. Pro­ rated mental age was significantly (p < .03) related to outcome in both groups, a finding that is consistent with reports from other investigators (DeMyer et al., 1981 ). In addition, abnormal speech was significantly (p < .0 l) related to outcome in Control Group l. Chronological age at onset of our treatment was not related to outcome, which is important because the two groups differed significantly on this variable at intake (by 6 months). The failure of CA to relate to outcome may be based on the very young age of all subjects at onset of treatment. Conceivably, a linear combination of pretreatment variables could have predicted outcome in the experimental group. Using a discriminant analysis (Ray, 1982) with the eight variables used in the first multivariate analysis, it was possible to predict perfectly the 9 subjects who did achieve normal functioning, and no subject was predicted to achieve this outcome who did not. In this analysis, PMA was the only variable that was sig­ nificantly related to outcome. Finally, when this prediction equation was applied to Control Group I subjects, 8 were pre­ dicted to achieve normal functioning with intensive treatment; this further verifies the similarity between the experimental group and Control Group I prior to treatment. Discussion This article reports the results of intensive behavioral treat· ment for young autistic children. Pretreatment measures re­ vealed no significant differences between the intensively treated experimental group and the minimally treated control groups. At follow-up, experimental group subjects did significantly bet­ ter than control group subjects. For example, 4 7% of the experi­ mental group achieved normal intellectual and educational functioning in contrast to only 2% of the control group subjects. The study incorporated certain methodological features de­ signed to increase confidence in the effectiveness of the experi­ mental group treatment: I. Pretreatment differences between the experimental and control groups were minimized in four ways. First, the assign­ ment of subjects to groups was as random as was ethically possi­ ble. The assignment apparently produced unbiased groups as evidenced by similar scores on the 20 pretreatment measures and by the prediction that an equal number of Control Group I and experimental group subjects would have achieved normal functioning had the former subjects received intensive treat­ ment. Second, the experimental group was not biased by receiv­ ing subjects with a favorable diagnosis or biased IQ testing be­ cause both diagnosis and IQ tests were constant across groups. Third, the referral process did not favor the project cases be­ cause there were no significant differences between Control Groups I and 2 at intake or follow-up, even though Control Group 2 subjects were referred to others by the same agency. 8 0. IVAR LOVAAS Fourth, subjects stayed within their groups, which preserved the original ( unbiased) group assignment. 2. A favorable outcome could have been caused not by the experimental treatment but by the attitudes and expectations of the staff. There are two findings that contradict this possibil­ ity of treatment agency (placebo) effects. First, because Control Group 2 subjects had no contact with the project, and because there was no difference between Control Groups I and 2 at fol­ low-up, placebo effects appear implausible. Second, the within­ subjects study showed that at least one treatment component contributed to the favorable outcome in the intensive treatment ( experimental) group. 3. It may be argued that the treatment worked because the subjects were not truly autistic. This is counterindicated by the high reliability of the independent diagnosis and by the out­ come data from the control groups, which are consistent with those reported by other investigators (Brown, 1969; DeMeyer et al., 1973; Eisenberg, 1956; Freeman et al., 1985; Havelkova, 1968; Rutter, 1970) for groups of young autistic children diag­ nosed by a variety of other agencies. 4. The spontaneous recovery rate among very young autistic children is unknown, and without a control group the favorable outcome in the experimental group could have been attributed to spontaneous recovery. However, the poor outcome in the sim­ ilarly constituted Control Groups I and 2 would seem to elimi­ nate spontaneous recovery as a contributing factor to the favor­ able outcome in the experimental group. The stability of the IQ test scores in the young autistic children examined by Freeman et al. (1985) attests once again to the chronicity of autistic be­ haviors and serves to further negate the effects of spontaneous recovery. 5. Posttreatment data showed that the effects of treatment (a) were substantial and easily detected, (b) were apparent on comprehensive, objective, and socially meaningful variables (IQ and school placement), and (c) were consistent with a very large body of prior research on the application of learning the­ ory to the treatment and education of developmentally disabled persons and with the very extensive ( 100-year-old) history of psychology laboratory work on learning processes in man and animals. In short, the favorable outcome reported for the inten­ sive-treatment experimental group can in all likelihood be at­ tributed to treatment. A number of measurement problems remain to be solved. For example, play, communicative speech, and IQ scores define the characteristics of autistic children and are considered pre­ dictors of outcome. Yet the measurement of these variables is no easy task. Consider play. First, play undoubtedly varies with the kinds of toys provided. Second, it is difficult to distinguish low levels of toy play (simple and repetitive play associated with young, normal children) from high levels of self-stimulatory be­ havior (a psychotic attribute associated with autistic children). Such problems introduce variability that needs immediate at­ tention before research can proceed in a meaningful manner. The term normalfunctioning has been used to describe chil­ dren who successfully passed normal first grade and achieved an average IQ on the WISC-R. But questions can be asked about whether these children truly recovered from autism. On the one hand, educational placement is a particularly valuable measure of progress because it is sensitive to both educational accom­ plishments and social-emotional functions. Also, continual promotion from grade to grade is made not by one particular teacher but by several teachers. School personnel describe these children as indistinguishable from their normal friends. On the other hand, certain residual deficits may remain in the normal functioning group that cannot be detected by teachers and par­ ents and can only be isolated on closer psychological assess­ ment, particularly as these children grow older. Answers to such questions will soon be forthcoming in a more comprehensive follow-up (McEachin, 1987). Several questions about treatment remain. It is unlikely that a therapist or investigator could replicate our treatment pro­ gram for the experimental group without prior extensive theo­ retical and supervised practical experience in one-to-one be­ havioral treatment with developmentally disabled clients as de­ scribed here and without demonstrated effectiveness in teaching complex behavioral repertoires as in imitative behavior and ab­ stract language. In the within-subjects studies that were re­ ported, contingent aversives were isolated as one significant variable. It is therefore unlikely that treatment effects could be replicated without this component. Many treatment variables are left unexplored, such as the effect of normal peers. Further­ more, the successful mainstreaming of a 2-4-year-old into a normal preschool group is much easier than the mainstreaming of an older autistic child into the primary grades. This last point underscores the importance of early intervention and places limits on the generalization of our data to older autistic chil­ dren. Historically, psychodynamic theory has maintained a strong influence on research and treatment with autistic children, offering some hope for recovery through experiential manipula­ tions. By the mid- I 960s, an increasing number of studies re­ ported that psychodynamic practitioners were unable to deliver on that promise (Rimland, 1964 ). One reaction to those failures was an emphasis on organic theories of autism that offered little or no hope for major improvements through psychological and educational interventions. In a comprehensive review of re­ search on autism, DeMyer et al. ( 1981) concluded that "[in the past) psychotic children were believed to be potentially capable of normal functioning in virtually all areas of development . . . during the decade of the 1970s it was the rare investigator who even gave lip-service to such previously held notions ... infan­ tile autism is a type of developmental disorder accompanied by severe and, to a large extent, permanent intellectual/behavioral deficits" (p. 432). The following points can now be made. First, at least two distinctively different groups emerged from the follow-up data in the experimental group. Perhaps this finding implies differ­ ent etiologies. If so, future theories of autism will have to iden­ tify these groups of children. Second, on the basis of testing to date, the recovered children show no permanent intellectual or behavioral deficits and their language appears normal, contrary to the position that many have postulated (Rutter, 197 4; Chur­ chill, 1978) but consistent with Kanner's (1943) position that autistic children possess potentially normal or superior intelli­ gence. Third, at intake, all subjects evidenced deficiencies across a wide range of behaviors, and during treatment they showed a broad improvement across all observed behaviors. The kind of(hypothesized) neural damage that mediates a par­ ticular kind of behavior, such as language (Rutter, 1974), is not consistent with these data. 9 TREATMENT OF AUTISTIC CHILDREN Although serious problems remain for exactly defining au­ tism or identifying its etiology, one encouraging conclusion can be stated: Given a group of children who show the kinds of be­ havioral deficits and excesses evident in our pretreatment mea­ sures, such children will continue to manifest similar severe psychological handicaps later in life unless subjected to inten­ sive behavioral treatment that can indeed significantly alter that outcome. These data promise a major reduction in the emotional hard­ ships of families with autistic children. The treatment proce­ dures described here may also prove equally effective with other childhood disorders, such as childhood schizophrenia. Certain important, practical implications in these findings may also be noted. The treatment schedule of subjects who achieved normal functioning could be reduced from 40 hr per week to infrequent visits even after the first 2 years of treatment. The assignment of one full-time special-education teacher for 2 years would cost an estimated $40,000, in contrast to the nearly $2 million in­ curred (in direct costs alone) by each client requiring life-long institutionalization. References Ackerman, A. B. ( 1980). The contribution of punishment to the treat­ ment of preschool aged children. Unpublished doctoral dissertation, University of California, Los Angeles. American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author. Bayley, N. ( 19 5 5 ). On the growth of intelligence. American Psychologist, JO. 805-818. Brecht, M. L., & Woodward, J. A. ( 1984 ). GANOVA: A univariate/multi­ variate analysis of variance program for the personal computer. Edu­ cational and Psychological Measurement. 44, 169-173. Brown, J. ( 1969). Adolescent development of children with infantile psychosis. Seminars in Psychiatry. 1, 79-89. Cattell, P. ( 1960). The measurement of intelligence of infants and young children. New York: Psychological Corporation. Churchill, D. W. (1978). Language: The problem beyond conditioning. In M. Rutter & E. Schopler (Eds.), Autism: A reappraisal of concepts and treatment (pp. 71-85). New York: Plenum. DeMyer, M. K., Barton, S., DeMyer, W. E., Norton, J. A., Allen, J., & Steele, R. ( 1973). Prognosis in autism: A follow-up study. Journal of Autism and Childhood Schizophrenia, 3. 199-246. DeMyer, M. K., Hingtgen, J. N., & Jackson, R. K. (1981). Infantile autism reviewed: A decade of research. Schizophrenia Bulletin. 7, 388-451. Doll, E. A. (1953). The measurement of social competence. Minneapo­ lis, MN: Minneapolis Educational Test Bureau. Dunn, L. M. (1981). Peabody Picture Vocabulary Test. Circle River, MI: American Guidance Service. Education for All Handicapped Children Act of 1975. Washington, DC: Congressional Record. Eisenberg, L. ( 1956). The autistic child in adolescence. American Jour­ nal of Psychiatry, 112, 607-612. Freeman, B. J., Ritvo, E. R., Needleman, R .. & Yokota, A. ( 1985). The stability of cognitive and linguistic parameters in autism: A 5-year study. Journal o(the American Academy of Child Psychiatry. 24. 290- 31 I. Gesell, A. ( 1949). Gesell Developmental Schedules. New York: Psycho­ logic-al Corporation. Havelkova, M. ( 1968). Follow-up study of7 l children diagnosed as psy­ chotic in preschool age. American Journal of Orthopsychiatry. 38, 846-857. Kanner, L. (1943). Autistic disturbances of affective contact. Nervous Child, 2, 217-250. Leiter, R. G. ( 1959). Part I of the manual for the 1948 revision of the Leiter International Performance Scale: Evidence of the reliability and validity of the Leiter tests. Psychology Service Center Journal, Jl,J-72. Lotter, V. (1967). Epidemiology ofautistic conditions in young children: II. Some characteristics of the parents and children. Social Psychia­ try. l, 163-173. Lovaas, 0. I., Ackerman, A. B., Alexander, D .. Firestone. P .. Perkins, J., & Young, D. (1980). Teaching developmentally disabled children: The me book. Austin, TX: Pro-Ed. Lovaas, 0. I., Koegel, R. L.. Simmons, J. Q., & Long, J. (1973). Some generalization and follow-up measures on autistic children in behav­ ior therapy. Journal of Applied Behavior Analysis, 6, 131-166. McEachin, J. J. (1987). Outcome of autistic children receiving intensive behavioral treatment. Residual deficits. Unpublished doctoral disser­ tation, University of California, Los Angeles. McEachin, J. J., & Leaf, R. B. (1984, May). The role of punishment in motivation of autistic children. Paper presented at the convention of the Association for Behavior Analysis, Nashville, TN. Ray, A. A. (1982). Statistical Analysis System user's guide. Statistics, 1982 edition. Cary, NC: SAS Institute. Rimland, B. (1964). Infantile autism. New York: Appleton-Century­ Crofts. Rutter, M. ( 1970). Autistic children: Infancy to adulthood. Seminars in Psychiatry. 2. 435-450. Rutter, M. ( 1974). The development of infantile autism. Psycholog1cal Medicine, 4, 147-163. Stutsman, R. ( 1948). Guide for administering the Merrill-Palmer Scale of Mental Tests. New York: Harcourt, Brace & World. Thorndike, R. L. ( I 972). Manua/for Stanford-Binet Intelligence Scale. Boston: Houghton Mifflin. Wechsler, D. ( 1967). Manualfor the Wechsler Pre-School and Primary Scale of Intelligence. New York: Psychological Corporation. Wechsler: D. (1974). Manual for the Wechsler Intelligence Scale for Children-Revised. New York: Psychological Corporation. Received October IO, 1985 Revision received March 28, 1986 • -research paper writing service