Changes in Language Development Among Autistic and Peer Children in Segregated and Integrated Preschool Settings

Sandra L. Harris, Jan S. Handleman, Barbara Kristoff, Leisa Bass, and Rita Gordon
Rutgers. The State University of New Jersey

Five young children with autism enrolled in a segregated class, five other children with autism in an integrated class, and four normally developing peer children in the integrated class were compared for developmental changes in language ability as measured by the Preschool Language Scale before and after training. The results, based on Mann-Whitney U tests, showed that (a) all of the children as a group made better than normative progress in rate of language development, (b) the scores of the autistic children were significantly lower than the peers before and after treatment, and (c) there were no significant differences in changes in language ability between the autistic children in the segregated and integrated classes.

In recent years an exciting body of research has emerged documenting that young children with autism may benefit socially (e.g., Odom, Hoyson, Jamieson, & Strain, 1985; Odom & Strain, 1986; Strain, Hoyson, & Jamieson, 1985) and educationally (e.g., Hoyson, Jamieson, & Strain, 1984) from being in a class with normally developing peers. The most thoroughly documented of these programs, the LEAP model at the University of Pittsburgh, has also shown that normal peers benefit academically from this educational experience (e.g., Hoyson et al., 1984).

This study explored the extent to which the language development of young children with autism might be differentially influenced by being in a segregated versus integrated classroom and examined the benefits to language development for a group of normal peers in an integrated class.

The opening of two new preschool classes for children with autism at the Douglass Developmental Disabilities Center (DDDC) in 1987, one integrated and one segregated, gave us an unusual opportunity to study the deficits of these two environments on children's language. Although the assignment of the autistic children in the two classes was not random, but was done on the basis of the severity of the child's behavior problems such as aggression, tantrums, and self stimulation, in other respects, including chronological age (CA), IQ, and language development the groups were comparable. These important similarities permitted us to use the two groups of children in a quasi-experimental design. Given the ethical and educational constraints on true random assignment, this fortunate placement of two groups of children whose intellectual and verbal skills were quite similar, in classes which differed primarily in terms of the greater frequency of significant behavior problems in the segregated class, provided an unusual opportunity to examine an interesting question.

Based on the literature suggesting the potential value of peer models for the educational and social progress of children with autism we predicted that (a) all children, both autistic and normal, would benefit from an intensive language curriculum and would show faster rates of learning after treatment than they had before; (b) autistic children in the integrated class would make more progress in language development than children in the segregated class; and (c) normal peers would develop language at a faster rate than the children with autism.

METHOD

Subjects

The 14 children whose data are presented here represent three groups of youngsters: 5 autistic children in a segregated preschool class, 5 autistic children in an integrated preschool class, and 4 normal peers in the integrated class (Table I). The autistic children were diagnosed as exhibiting Infantile Autism (American Psychiatric Association, 1980) by both the referring agency and an experienced clinical psychologist (S.L.H.) at our Center. As a group, these autistic children should be characterized as relatively high functioning in contrast to the total population of children with autism. There were no children with known seizure disorders in either preschool class, and young children who showed more profound disability than these 10 youngsters were placed in other classes in our Center.

Language Development in Integrated and Segregated Classes

Segregated Class. The mean CA of the four boys and one girl in the segregated class at the time of initial testing was 58 months (range 49-66 months). Their mean IQ on the Stanford-Binet was 62.6 (range 43-71). The Peabody Picture Vocabulary Test Revised (PPVT-R; Dunn & Dunn, 1981) yielded a similar pattern with a group mean of 68.2 (range 46-97). On the Childhood Autism Rating Scale (CARS; Schopler, Reichler, DeVellis, & Daly 1980) the mean was 34 (range 30.5-38); two of the children were in the severely autistic range of the CARS and three in the mild to moderate category.

Integrated Class — Autistic Children. There were four boys and one girl with autism in the integrated class. Their mean age was 55.8 months (range 52-60 months). On the Stanford-Binet they earned a mean IQ of 69 (range 59-83) and on the PPVT-R the mean was 60.4 (range 41-90). The mean CARS score was 30.4 (range 23.5-38); two of the children in this group were classified as non-autistic on the CARS at the time they entered the integrated class. They had been in a segregated class at the Center for 1 year when the CARS was administered and we have no pretreatment CARS data for them. These children had however each been diagnosed as autistic according to the DSM-III criteria (American Psychiatric Association, 1980) at the time of admission. Among the three remaining children, two were in the mild to moderate category of autism on the CARS and one in the severe category.

Integrated Class-Normal Peers. The normal peers included two boys and two girls whose mean age was 45.25 months (range 40-52 months) at initial testing. They were recruited through word of mouth and a newspaper ad. One was the daughter of a staff member and one the brother of a child with autism in another component of the Center's program; the other three children lived in the local community. We have excluded from this report the data of one boy who spoke only Chinese at the intake and whose data would therefore not have been meaningful. The mean Stanford-Binet IQ of the four peers was 108.5 (range 96-125) and the mean PPVT-R score was 97 (range 83-109). None of these children had any known learning or behavior problems and they were selected because of their social and interpersonal responsiveness to peers and adults. They all came from middle-class, intact homes.

Table I summarizes important demographic variables in the three groups. Two-tailed Mann-Whitney U tests (Siegel, 1956) comparing the two groups of autistic subjects on the variable of CA, U = 9.5, p = ns; CARS scores, U = 6, p = ns; Stanford-Binet IQ, U = 9, p = ns; and PPVT-R score, U = 9, p = ns, failed to reveal any significant differences between the two groups. The peer children were significantly younger, U = 2.5, p < .02, and earned significantly higher Stanford-Binet IQs, U = 0, p < .002, and PPVT-R scores, U = 2.75, p < .02 than the children with autism.

Harris, Handleman, Kristoff, Bass, and Gordon

Table I. Demographic Comparisons of Subject Groups

CA Binet

Group (months) IQ PPVT-R CARS

Segregated class 58 62.6 68.2 34.0

Integrated (autistic) 55.8 69.0 60.4 30.4

Integrated (peer) 45.25 108.5 97.0 NA

General Curriculum

The curriculum for the integrated and segregated classes is based upon the same set of goals and objectives. This curriculum, which is developmentally organized and language focused, includes the areas of receptive and expressive language, cognition (preacademic arithmetic and reading), fine motor, gross motor, socialization (play and peer interaction, affect and self-concept, adult interactions, classroom behavior), and self-help skills.

The instructional format consists of individualized programs, small group, and class lessons. A daily language group focuses on developmentally appropriate language and classroom readiness skills. The daily schedule is consistent with that of a typical preschool and draws upon incidental learning experiences as well as more structured teaching.

The experiences of the children in the two classes differed in that the segregated class provided more individualized instruction and their large group activities consisted of 5 children while the integrated class provided less individualized attention with instruction in small (3 to 4) and large (10 children) groups. The adult child ratio was 3:5 in the segregated class and 3:l0 in the integrated class.

All of the children in both classes were exposed to formal, structured group language instruction in the classroom including a weekly group led by the speech and language specialist. Individual speech and language goals were developed by the speech therapist using converging data from the Preschool Language Scale (PLS; Zimmerman, Steiner, & Pond, 1979), Evaluating Skills in Communication (Riley, 1984), and a developmentally sequenced checklist that we have written at the DDDC.

Assessment

The PLS (Zimmerman et al., 1979) was administered to each child by a certified speech therapist during an initial assessment and 5 to 11 months later. The time from pre to posttesting varied as a function of the availability of staff to do testing, with the normal peers having the last priority. The PLS yields a measure of Language Ability (LA) based upon the subscales of Auditory Ability and Verbal Ability. The test provides an estimate of developmental status in months as well as a quotient similar to an IQ; we used test age in months for data analysis.

According to its creators, the PLS has a split-half reliability of .88; it correlates .97 with the Illinois Test of Psycholinguistic Ability, .59 with the PPVT, and .66 with the 1960 Stanford-Binet (Zimmerman et al., 1979). Zimmerman et al. (1979) also reported support for the predictive validity of the PLS with scores of 79% of children from a Head Start program predicting language status 1 year later. The normative sample for the PLS includes Head Start children, children in early childhood education programs, and youngsters from middle-class nursery schools.

Each child's developmental language status at pretesting was calculated by a formula that uses developmental age in months at pretesting divided by chronological age in months at pretesting (Wolery, 1983) as an estimate of rate of progress prior to training. Progress in the program was estimated with the Intervention Efficiency Index of developmental gain (Bagnato & Neiswoith, 1980). This is calculated by dividing number of months of developmental progress from pretest to posttest by total number of months in the program. Using this formula we were able to compensate for the variation in time from pre to posttesting for the children. For both formulas, if a child progresses 1 month developmentally for each month that passes, the index score would be 1.00. Both of these formulas were used by Hoyson et al. (1984) to assess the progress of children in their integrated preschool. We refer to data from these formulas when describing rate of progress as opposed to developmental months. We have included both rate and developmental age because these two measures provide somewhat different perspectives on the child's growth.

RESULTS

Mann-Whitney U tests (Siegel, 1956) of LA scores were used to compare the differences between groups. In those instances where the direction of differences had been predicted, we used one-tailed tests and in all other cases, two-tailed tests.

Table II includes the pre- and posttraining developmental scores in months and rate of development for each child on the LA measure of the PLS. This table also summarizes the group mean scores.

Rate of Development

A Mann-Whitney U test comparing the rates of developmental progress of all 14 subjects before (M = 0.83) and after (M = 1.63) treatment reveals a highly significant increase in their rate of progress, U = 24, p < .001 (one-tailed).

Table II. Developmental Age and Rate of Progress Pre and Posttraining on the PLS

Pretraining Posttraining

CA LA LA/CA CA LA LA/CA

Child (months) (months) (rate) (months) (months) (rate)

Segregated class

SI 60 49.50 0.83 68 57.75 1.03

S2 60 42.00 0.70 66 49.50 1.25

S3 55 36.75 0.67 61 48.75 2.00

S4 66 26.25 0.40 74 39.75 1.69

S5 49 35.25 0.72 55 40.50 0.88

Class mean 37.95 0.66 47.25 1.37

Integrated classAutistic children

I1 52 39.75 0.76 63 42.75 0.27

12 60 29.25 0.49 66 46.50 2.88

I3 55 36.00 0.66 61 45.75 1.63

I4 52 40.50 0.78 63 54.75 1.30

15 60 56.25 0.94 71 72.75 1.50

Class mean 40.35 0.73 52.50 1.52

Integrated class Peers

IPI 40 57.75 1.44 45 66.75 ! .80

IP2 49 55.50 1.11 54 64.50 1.80

IP3 52 51.00 0.98 59 65.25 2.04

IP4 40 43.50 1.09 45 57.00 2.70

Class mean 51.90 1.15 63.37 2.09

A Mann-Whitney U test comparing the rates of developmental progress of all 14 subjects before (M = 0.83) and after (M = 1.63) treatment reveals

Consistent with the finding for all subjects, the children in the segregated class progressed at a more rapid rate after training than they had before, U = 0, p < .004 (one-tailed). Their pretreatment mean developmental rate of language ability was 0.66; after training this had risen to a mean of 1.37. All of the children showed progress on this measure.

The autistic children in the integrated class showed a strong trend toward measurable progress in rate of LA, U = 5,p < .075 (one-tailed). Their group mean before treatment was 0.73; after treatment this mean rose to 1.52. Only subject I-1 failed to show measurable progress.

The normal peers in the integrated class had the highest initial developmental rate and made even more rapid progress after training, U = 0, p < .014. Their pretreatment mean was 1.15; at posttesting the mean was 2.09. Each of these children did better after training than before.

A Mann-Whitney U test comparing the developmental rate on LA of all autistic children (M = 0.70) before treatment with their normal peers before treatment (M = 1.15) indicates a highly significant difference between the group, U = 0, p < .001 (one-tailed). After training there remained a trend toward a significant difference between the posttraining rate of the autistic (M =1.44) and peer (M = 2.09) children, U = 6, p < . 10 (two-tailed).

Developmental Age

A comparison of developmental age, as opposed to rate of development, indicates that all of the children as a group had higher scores after training (M ' = 53.7 months) than they did before (M = 42.8 months), U = 46, p < .01 (one-tailed). Before training the autistic children had significantly lower age scores (M = 39.15) than the peers (M = 51.9), U = 4, p < .05 (one-tailed). This held true after training as well, with the autistic children continuing to score lower in months (M = 49.8) than the peers (M = 63.37), U = 5, p < .05 (two-tailed).

The peer children (3M = 63.37), U = 1, p = .029, and the autistic children in the integrated class (M = 52.5), U = 4, p < .048 both were functioning at a significantly higher developmental level in months after training than they had before (peers, M = 51.9; autistic, M = 40.35). The autistic children in the segregated class showed a trend in this same direction (pre M = 37.95, post (M = 47.25), U = 5.5, p < .111.

Segregated Versus Integrated

Inspection of the group means for developmental level at pre and posttesting reveals that the children in the segregated and integrated classes started at approximately the same level of language ability and remained close after training. A Mann-Whitney U test of developmental level in months indicates no significant differences between the two groups before, U I0, p = ns, or after training, U = I0, p = ns. A parallel analysis for rate of development likewise reveals no significant differences between the segregated and integrated groups before, U = 10, p = ns, or after training, U = 11, p ns.

DISCUSSION

The results of the present study document substantial changes in language abilities of preschool children with autism who are exposed to an intensive language stimulation program. It also supports the benefits for normal peers of involvement in an integrated preschool class.

Both groups of autistic children increased their developmental language rate from a pace that fell below that of 1 month's progress for each month of chronological age to a rate that more nearly approximated normal development. In fact, their posttraining rates of development resemble those of the normal peers at intake. The peers were nonetheless developmentally more advanced after training with a mean group language ability in months significantly exceeding that of the autistic children. The peers' language ability at posttesting was nearly 13 months ahead of their CA, while the autistic children were 15 months behind their CA. Although that 15 months discrepancy is of considerable clinical concern, at the time of the pretest the gap between CA and LA for the autistic children had been 18 months, and thus had been narrowed by 3 months in an average of about 8 months. The autistic children were gaining ground rather than falling farther behind.

Only one child of the 14, 1-1 showed a decline in rate of progress on the PLS. All of the other children either maintained a steady rate of progress or showed an increase in their rate of development over time. Because I-1 has not responded well to formal testing throughout his career with us, it is difficult to know to what extent his apparent lack of progress reflects our difficulties in measurement as opposed to his failure to acquire new skills. It is however noteworthy that he was integrated into a public school program for children who are neurologically impaired and has made a good adjustment to that setting. That observation coupled with his consistently satisfactory progress on individual educational objectives in the classroom suggests that the standardized testing may not have been a useful reflection of his progress.

Although the failure to find differences between the integrated and segregated classes might be taken to support an argument that all autistic children should be placed in integrated classes and hence in more normalized and less expensive settings, we do not believe that is an appropriate conclusion at this time. Our primary reservation is that the children in the segregated class posed substantial management problems that required individual attention and, in our judgment, had the potential to be disruptive to class functioning in a less intensively staffed class. Whether these clinical concerns would be borne out in operation is of course an important empirical question.

Although the integrated class was created to facilitate the development of children with autism, it is gratifying and important to note that the normal peers made striking progress as well. These children, all of whom were progressing at an appropriate developmental rate prior to enrollment, were not held back in development by the presence of their handicapped peers and continued to show growth in language skills.

It is of interest to note that there do not appear to be differences between the children in the integrated and segregated classes in terms of their rates of language development over the course of this instruction. Although it is well documented that peers are valuable as social, interpersonal, models of behavior (e.g., Beckman & Kohl, 1987; Odom et al., 1985; Odom & Strain, 1986), their contributions to the language development of children with autism may not be as significant or may have to be manipulated more systematically than we have done to date. Alternatively, the changes facilitated by peers may be reflected in domains other than those tapped by the PLS. It is important to examine this question in detail in future research. At present it can only be said that with chronological and language ages that were initially similar, these two groups of children made parallel progress on the PLS.

In sum, our results provide strong support for the value of early, intensive language intervention for increasing the developmental pace of relatively high functioning children with autism and their normally developing peers. The contributions of the peers to the language acquisition of the children with autism remains to be explored in greater detail.

REFERENCES