Effects of differing levels of inclusion on preschoolers with disabilities.

The practice of inclusion (i.e., educating children with and without disabilities together) derives from the principle of least restrictive environment and the civil rights movement. Although some parents and professionals call for universal full inclusion (Peck, 1995), others advocate For the maintenance of a continuum of services (Bricker, 1995; Strain, 1995), suggesting that periods of special education-only instruction may at times be beneficial (Fuchs & Fuchs, 1994). However, most current research on service delivery has shifted from questions about the relative efficacy of inclusion to questions about effective inclusion models (Guralnick, 1990; Salisbury & Vincent, 1990). The present study represents a return to the earlier concern, the efficacy of varying degrees of inclusion.

The key issue in this debate is whether the individual needs of some children may be lost when a single approach is applied universally. Because children with disabilities form all extraordinarily heterogeneous population, there are risks in adopting a "one size fits all" philosophy across different subgroups of this population. For example, in a multisite study of inclusion models for elementary-age children with learning disabilities, approximately half of the children showed little or no growth in reading achievement (Zigmond, Jenkins, Fuchs, Deno, & Fuchs, 1995; Zigmond, Jenkins, Fuchs, Deno, Fuchs, Baker, et al., 1995). In contrast, children with more severe disabilities were found to generally benefit from inclusion (Buysse & Bailey, 1993), although the generality of this conclusion has been challenged by MacMillan, Gresham, and Forness (1996). Because preschool children with disabilities also constitute a diverse population, varying broadly in social, linguistic, cognitive, and motor skills as well as in other factors, individual children may respond differently to different educational environments. Thus, studies seeking to uncover best practices must take into account the characteristics of the research participants.

Most research on special education preschool programs has concentrated on group effects, providing information about the average benefit of a practice such as inclusion (Buysse & Bailey, 1993; Lamorey & Bricker, 1993; Odom & McEvoy, 1988). However, there is evidence that subgroups of children may not respond similarly to the same approach. In a previous study involving random assignment to either special education-only classes or integrated classes (i.e., four children developing typically and eight children who qualified for special education services), Cole, Mills, Dale, and Jenkins (1991) found no main effect differences for children in either model, but significant Aptitude X Treatment interactions (ATIs). Specifically, children with disabilities who performed relatively higher on pretest measures benefited more from integrated special education classrooms, while children who performed lower on pretest measures benefited more from special education-only classrooms. These findings suggest that service delivery models may relate to developmental outcomes in more complicated ways than first thought (i.e., interactions between the type of service delivery arrangements and child characteristics). It is possible that the findings of no differences in developmental measures in earlier studies of mainstreaming may have masked ATIs.

If some young children are better served academically in special education-only settings and others are better served in integrated special education settings, then a continuum of service delivery options is preferable to any single approach. Given our earlier findings, we felt it important to seek replication. We also wanted to examine the relative effects of a third, even more inclusive model of integration, one in which typically developing children are in the majority. Thus, the present study examined the effects of three different levels of integration--special education-only (classrooms which enroll only children with disabilities), integrated special education (more children with disabilities than typically developing children), and mainstreamed (more typically developing children than children with disabilities)--on developmental outcomes of young children with disabilities.

METHOD

Subjects

A total of 66 children, enrolled in a laboratory school in the United States, participated in this study. These children, ages 31 months to 75 months (SD = 11.23), qualified for special education under the state's noncategorical system of funding for preschool children. Children qualify for special education if they exhibit a delay of at least 1.5 standard deviations on a normed test in two or more developmental areas (gross motor, fine motor, language, cognition, or social-emotional development) or if they exhibit a delay of 2 or more standard deviations in one of these areas. Forty-six of the students exhibited a significant delay in gross motor (70%), 55 in fine motor (83%), 55 in language (83%), 36 in cognition (55%), and 56 in social-emotional development (85%). One-way ANOVAs on each area of delay by level of integration revealed no significant differences. The sample contained 49 boys (74%) and 17 girls (26%), with an ethnic mix of 45 European American (68%); 13 African American (20%); and 8 Asian American, Native American, Pacific Islander American, or other (12%). See Table 1 for a demographic breakdown of the sample.

TABLE 1 Description of Subjects/hr Each Level of Integration

                        Special Education-
Descriptive Variable          Only

n                              22
Age                       51.77 months
Range                     36-72 months

                           n       %
Gender
Boys
Girls                      5       23
Ethnicity
European American         14       64
 African American          4       18
Other                      4       18

                        Integrated Special
                            Education

n                              22
Age                       52.05 months
Range                     36-70 months

                          n        %
Gender
Boys
Girls                     6       27
Ethnicity
European American        13       60
 African American         7       31
Other                     2        9

                         Mainstreamed

n                              22
Age                       55.95 months
Range                     31-75 months

                          n        %
Gender
Boys                      16      73
Girls                      6      27
Ethnicity
European American         18      82
 African American          2       9
Other                      2       9

In addition, 51 typically developing children were enrolled in integrated and mainstreamed classrooms. These children, ages 32 months to 65 months (M = 47 months, SD = 7.88), included 31 boys (61%) and 20 girls (39%), representing ail ethnic mix of 39 European American (76%); 9 African American (18%); and 3 Asian, Native American, Pacific Islander American, or other (6%). For this group the McCarthy General Cognitive Index (GCI; McCarthy, 1972) mean was 105 (SD = 15.28) and the Preschool Language Assessment Instrument (PLAI; Blank, Rose, & Berlin, 1978) mean proportion for totally appropriate responses was .62 (SD = .20). Only data from children with disabilities are reported in the remainder of this article.

Design

Children were randomly assigned to one of three levels of inclusion (special education-only, integrated special education, and mainstreamed classrooms). Each classroom enrolled 14 children. Special education-only classrooms enrolled 14 children with disabilities and no typically developing children. The integrated classrooms enrolled 3 typically developing children and 11 children with disabilities. Mainstreamed classrooms enrolled 9 typically developing children and 5 children with disabilities.

Final Sample Selection. The original sample included 114 children with disabilities, but some of these had participated in the preschool program for more than 1 year. To ensure comparability, we used only data from a child's first year in the program. Moreover, the number of children with disabilities who had been assigned to the three types of treatment ranged from 22 in the mainstreamed condition to 48 in the special education-only condition. Because differential weighting from disproportionate samples creates problems in interpreting interaction terms, we equated sample size by matching a subsample of children from the integrated special education and special education-only programs (n = 22 each) with students in the mainstreaming condition, using the PLAI pretest scores for total appropriate responses as the matching variable. As shown in Table 2, the three groups were similar at pretest, with no significant differences among the final groups on any pretest measure. Even though we had randomly assigned children to classrooms and conditions, our final selection of students matched for pretest PLAI performance resulted in a quasi-experimental design. However, a comparison of results found for the larger sample mirrored those for the smaller sample reported here. Table 1 gives descriptive statistics for each classroom composition.

TABLE 2 Repeated Measures ANOVAs on McCarthy Scales and PLAI

                                 Pretest
Measure                        M      (SD)

McCarthy GCI
 Special education - only    67.05   (19.78)
 Integrated                  67.73   (20.26)
 Mainstreamed                64.91   (16.52)
 Time
 Interaction
McCarthy Verbal
 Special education - only    33.95   (10.69)
 Integrated                  31.45   (11.22)
 Mainstreamed                30.09    (9.84)
 Time
 Interaction
McCarthy Perceptual
 Segregated                  31.57   (11.24)
 Integrated                  33.68   (11.89)
 Mainstreamed                33.48   (10.15)
 Time
 Interaction
McCarthy Quantitative
 Special education - only    33.91    (9.59)
 Integrated                  35.50   (12.73)
 Mainstreamed                32.00    (9.19)
 Time
 Interaction
McCarthy Memory
 Segregated                  34.38   (11.27)
 Integrated                  34.64   (10.44)
 Mainstreamed                32.38    (9.21)
 Time
 Interaction

PLAI Total Appropriate
 Special education - only      .31     (.27)
 Integrated                    .33     (.26)
 Mainstreamed                  .31     (.26)
 Time
 Interaction

                                 Posttest
Measure                        M      (SD)

McCarthy GCI
 Special education - only    74.14   (28.32)
 Integrated                  75.18   (23.65)
 Mainstreamed                65.33   (16.16)
 Time
 Interaction
McCarthy Verbal
 Special education - only    34.33   (12.20)
 Integrated                  36.41   (13.05)
 Mainstreamed                30.14    (9.60)
 Time
 Interaction
McCarthy Perceptual
 Segregated                  33.19   (12.05)
 Integrated                  36.36   (14.47)
 Mainstreamed                33.76    (9.54)
 Time
 Interaction
McCarthy Quantitative
 Special education - only    32.91   (11.39)
 Integrated                  37.82   (13.85)
 Mainstreamed                32.95    (9.89)
 Time
 Interaction
McCarthy Memory
 Segregated                  34.05   (13.19)
 Integrated                  37.64   (15.65)
 Mainstreamed                30.00   (10.20)
 Time
 Interaction

PLAI Total Appropriate
 Special education - only      .44     (.20)
 Integrated                    .44     (.24)
 Mainstreamed                  .37     (.25)
 Time
 Interaction

Measure                      Significance

McCarthy GCI
 Special education - only
 Integrated
 Mainstreamed
 Time                        F = 6.01 (1, 61), p < .02
 Interaction                 F = 1.25 (2, 61), ns
McCarthy Verbal
 Special education - only
 Integrated
 Mainstreamed
 Time                        F = 3.47 (1, 61), ns
 Interaction                 F = 2.74 (2, 61), ns
McCarthy Perceptual
 Segregated
 Integrated
 Mainstreamed
 Time                        F = 3.49 (1, 61), ns
 Interaction                 F = .72 (1, 61), ns
McCarthy Quantitative
 Special education - only
 Integrated
 Mainstreamed
 Time                        F = .64 (1, 61), ns
 Interaction                 F = 1.05 (1, 61), ,s
McCarthy Memory
 Segregated
 Integrated
 Mainstreamed
 Time                        F = .01 (1,61),ns
 Interaction                 F = 2.23 (1, 61), ns

PLAI Total Appropriate
 Special education - only
 Integrated
 Mainstreamed
 Time                        F = 31.81 (1, 62), p < .000
 Interaction                 F = 1.58 (2, 62), ns

Note: McCarthy GCI = McCarthy Scales of Children's Abilities General Cognitive Index: PLAI = Preschool Language Assessment Instrument.

Data collection extended over a 3-year period, with all three types of classrooms represented each year. Children attended preschool for 2 hr and 15 min per day, 5 days per week, for 180 school days. There were morning and afternoon sessions for each of the three classroom compositions. Across the 3 years, four different teachers taught the mainstreamed, five taught integrated, and four taught special education-only classrooms.

Curricula

In the first 2 years of the project each classroom used one of two curricula. Five of the six classrooms used the Mediated Learning (ML) Curriculum (Osborn, Sherwood, & Cole, 1991) and one of the special education-only classrooms used the Direct Instruction (DI) program as developed by Engelmann and his colleagues (Becker, 1977; Becker, Engelmann, & Thomas, 1975). Due to the implementation of another research project in the third year, the curricula used in the classrooms changed. At that time one classroom representing each level of classroom composition employed direct language teaching methods (Waryas & Stremel-Campbell, 1984), with the other three classrooms employing an interactive curriculum (Drummond, 1989). One-way ANOVAs on posttest scores for children enrolled in each curriculum revealed no significant differences. Treatment fidelities were consistently high for each curricular approach; details of those analyses are available in Cole, Dale, Mills, and Jenkins (1993) and Cole, Mills, Dale, and Jenkins (1996).

Each classroom was staffed by a head teacher with a Master's degree in Special Education, an assistant teacher, a practicum student, and related service personnel who provided services in the classroom during segments of the school day. On average, three adults were in each classroom.

Measures

McCarthy Scales of Children's Abilities (MSCA). The MSCA (McCarthy, 1972) is an individually administered intelligence test for children ranging in age from 2 1/2 to 8 1/2 years of age. Subtests include Verbal, Perceptual, Quantitative, Memory, Motor, and the General Cognitive Index (GCI), composed of a combined set of the subsets. The MSCA is widely used and covers the appropriate developmental range for preschoolers with developmental delays. The following average split-half reliabilities are indicated: GCI = .93, Verbal = .88, Perceptual = .84, Quantitative = .81, and Memory = .79. Test-retest reliability ranges from .75 to .91. Sattler (1988) characterized it as "well standardized and psychometrically sound" (p. 295). We obtained complete pre- and posttesting results for 64 of 66 children on this measure.

Preschool Language Assessment Instrument (PLAI). The PLAI (Blank et al., 1978) is an experimental test designed to measure children's ability to respond to increasingly difficult and abstract language, similar to that encountered in teaching situations with four distinct ascending levels of abstraction: 1 Matching Perception (e.g., What do you see?); 2 Selective Analysis of Perception (e.g., What is different?); 3 Reordering Perception (e.g., How can you tell?); and 4 Reasoning About Perception (e.g., What will happen?). The split-half reliability for each level is: group 1 = .64, group 2 = .80, group 3 = .83, and group 4 = .86. Test-retest reliabilities for each group are: group 1 = .73, group 2 = .83, group 3 = .86, and group 4 = .88. A variety of scores can be derived from the test. For the present study, the total number of adequate responses (a scoring category defined in the test manual) was used as an overall measure. We obtained complete pre- and posttest results for 65 children on this test.

Testing Procedures

The McCarthy and PLAI were administered as pre- and posttests between October and May. The minimum period between pre- and posttests was 6 months. Research staff consisting of graduate students in speech language pathology, special education, and psychology as well as undergraduate students in psychology conducted all testing. Testing staff were not involved in classroom activities arid were not informed of the nature of the hypotheses involved in the study. Guidelines were in place to ensure that research staff tested approximately equal numbers of students across the three conditions.

RESULTS

Means and standard deviations for pre- and posttest measures of the three treatments are presented in Table 2. Analyses of variance (ANOVAs) compared the three groups' pretest levels on the following six measures: McCarthy GCI, Verbal, Perceptual, Quantitative, and Memory scale scores, arid PLAI number of total appropriate responses. There were no statistically significant pretest differences among the three groups.

Pre- and posttests were examined through repeated-measures ANOVAs, with treatment (i.e., class composition) as a between-subjects factor and time (pre- and posttest) as a within-subjects factor. The interaction of Treatment X Time, reflecting differences among treatment in amount of change, was not statistically significant for any of the six measures.

The main effect for time reached statistical significance on McCarthy GCI, F(1, 61), p [is less than] .05 and the PLAI, F(1, 62), p [is less than] .001. Recall that McCarthy scores are age-based standard scores; thus the same standard score on pre- and posttest indicates that children grew at the average of their norm-referenced group. The statistically significant pre- to posttest difference on the GCI resulted from participants' higher scores, relative to the norm group, at posttest than at pretest. The PLAI provides only raw scores (confounding maturation and education effects), so the overall pre- to posttest changes cannot be easily interpreted.

Exploring further the gains by treatment groups on the six measures, we computed effect sizes to quantify difference between pre- and posttest for each group (Glass, McGaw, & Smith, 1981). As shown in Table 3, the largest effect sizes were associated with the integrated treatment, followed by the special education-only treatment. The smallest effect sizes were associated with the mainstream treatment, for which all effect sizes from the McCarthy measures were either negative or near zero. Effect sizes observed for the PLAI, ranging from 60 to 78, were larger than those for the McCarthy, again reflecting a difference in the type of scores produced by the two tests (i.e., raw vs. standard scores). Contrast analyses were performed to evaluate the pre- to posttest change on each measure for the three groups separately. For the integrated treatment, differences between pre- to and posttest means were significant for McCarthy GCI (p = .036) and Verbal scale (p = .004), and nearly so for the Perceptual scale (p = .059). For the special education-only treatment, the pre- to posttest difference on the McCarthy GCI (p = .05) approached significance. For the mainstream treatment, no pre- to posttest differences approached significance on any McCarthy measure. All three treatments produced significant pre- to posttest changes on PLAI raw scores.

TABLE 3 Effect Sizes for gains from Pre- to Posttests for Three
Classroom Compositions

                                    Classroom Composition
Measure                    Special       Integrated   Mainstreamed
                        Education-Only
McCarthy
 GCI                          .36            .40           -.01
 Verbal                      -.01            .48           -.01
 Perceptual                   .17            .25            .02
 Memory                      -.10            .29           -.24
PLAI                          .78            .64            .60

Note: Effect sizes computed by dividing pre-posttest difference by the pretest standard deviation, pooled across groups. McCarthy GCI = McCarthy Scales of Children's General Cognitive Index; PLAI = Preschool Language Assessment Instrument.

We next examined ATIs for each posttest measure, using multiple regression analyses with hierarchical ordering as follows: age, pretest, class composition, and ability-(pretest cognitive and language scores)by-treatment. Ability pretests used in the regression were either PLAI total appropriate responses or McCarthy GCI. Because of the complexity of the ATIs, we conducted pairwise comparisons of the groups (R. Abbott, personal communication, April 28, 1992), that is, mainstreamed versus integrated special education, special education-only versus integrated special education, and special education-only versus mainstreamed. Results are summarized in Table 4. From the total of 36 regressions, 9 significant ATIs were observed.

TABLE 4 Summary of Aptitude x Treatment Multiple Regression Analysis for Students in Special Education-Only, Integrated Special Education, and Mainstreamed Settings

Pretest Measure   Posttest Measure   ISE vs. Mainstream

PLAI                 McCarthy
                       Verbal            p < .01(a)
                     [R.sup.2]              0.081
                     McCarthy
                       Quantitative      p < .05(a)
                     [R.sup.2]              0.046
                     McCarthy
                       Memory            p < .01(a)
                     [R.sup.2]              0.106
                     McCarthy
                       GCI               p < .05(a)
                     [R.sup.2]              0.055
                     McCarthy
                     Perceptual              ns
                     PLAI                    ns

McCarthy GCI         McCarthy
                       Verbal                ns
                     McCarthy
                       Quantitative             ns
                     [R.sup.2]               ns
                     McCarthy
                       Memory            p < .05(a)
                     [R.sup.2]              0.024
                     McCarthy
                       GCI                   ns
                     McCarthy
                       Perceptual            ns
                     PLAI                    ns

Pretest Measure   Posttest Measure   SE-O vs. ISE

PLAI                 McCarthy
                       Verbal            p < .01(a)
                     [R.sup.2]              0.068
                     McCarthy
                       Quantitative      p < .0.1(a)
                     [R.sup.2]              0.112
                     McCarthy
                       Memory            p < .01(a)
                     [R.sup.2]              0.095
                     McCarthy
                       GCI                   ns
                     [R.sup.2]
                     McCarthy
                       Perceptual
                     PLAI                    ns

McCarthy GCI         McCarthy
                       Verbal                ns
                     McCarthy
                       Quantitative      p < .05a
                     [R.sup.2]              0.039
                     McCarthy
                       Memory                ns
                     [R.sup.2]
                     McCarthy
                       GCI                   ns
                     McCarthy
                       Perceptual            ns
                     PLAI                    ns

Pretest Measure   Posttest Measure   SE-O vs. Maintream

PLAI                 McCarthy
                       Verbal                ns
                     [R.sup.2]
                     McCarthy
                       Quantitative          ns
                     [R.sup.2]
                     McCarthy
                       Memory                ns
                     [R.sup.2]
                     McCarthy
                       GCI                   ns
                     [R.sup.2]
                     McCarthy
                       Perceptual            ns
                     PLAI                    ns

McCarthy GCI         McCarthy
                       Verbal                ns
                     McCarthy
                       Quantitative          ns
                     [R.sup.2]
                     McCarthy
                       Memory                ns
                     [R.sup.2]
                     McCarthy
                       GCI                   ns
                     McCarthy
                       Perceptual            ns
                     PLAI                    ns

Note: ISE = integrated special education; SE-O = special education-only; McCarthy GCI = McCarthy Scales of Children's Abilities Genergal Cognitive Index; PLAI = Preschool Language Assessment Instrument.

(a) The direction of the slope is highest far ISE.

Special Education-Only Versus Integrated Special Education. After controlling for age and pretest, the interaction of pretest PLAI scores and treatment significantly predicted three posttests: McCarthy Verbal (p [is less than] .01), Quantitative (p [is less than] .01), and Memory (p [is less than] .01) posttest scale scores. The interaction between McCarthy GCI pretest and treatment also predicted McCarthy Quantitative (p [is less than] .05) posttest scale scores. The direction of the interaction for all four ATIs indicated that higher performing students at pretest gained more from the integrated special education classes, whereas lower performing students gained more from the special education-only classes. Figure 1 illustrates the pattern of all interactions.

[Figure 1 ILLUSTRATION OMITTED]

Integrated Special Education Versus Mainstreamed. Again, after controlling for age and pretest, the integrated versus mainstreamed treatment significantly interacted with pretest PLAI scores in predicting four posttests: McCarthy Verbal (p [is less than] .01), Quantitative (p [is less than] .05), Memory (p [is less than] .01), and GCI (p [is less than] .05) scale scores. The interaction between McCarthy pretest GCI and treatment was also a significant predictor of posttest McCarthy Memory (p [is less than] .05). The direction of the interactions was the same for all five ATIs: higher performing students at pretest gained more in integrated special education and lower performing students gained more in the mainstreamed classes.

Special Education-Only Versus Mainstreamed. No significant interactions with pretest ability measures were found.

We used the Johnson-Neyman procedure (Pedhazur, 1982) to determine the region of nonsignificance for each pair of regression lines. The statistically significant differences for lower performing students noted on the nine interactions of pretests and treatments did not occur within the range of scores observed in our sample. The differences between the higher performing students in the integrated special education relative to the special education-only or mainstreamed classrooms were, however, significant. Thus, while the ATIs are significant, their impact is greater for the higher performing children than for the lower performing children.

DISCUSSION

This study contrasted three classroom ratios of children with disabilities to typically developing peers. Across the total sample of children, we observed significant growth from pre- to posttest on cognitive (McCarthy GCI) and language (PLAI) measures. Although analyses of variance did not reveal significant treatment differences, within-group analyses of pre- to posttest gains indicated moderate effect sizes (.25 -- .48) for the integrated special education treatment, which was primarily responsible for the significant McCarthy gains. The special education-only treatment also produced a moderate level effect size (.36) on the McCarthy GCI. In contrast, effect sizes for the mainstream treatment were either negative or near zero across all McCarthy measures. All three treatments resulted in significant gains in language development, as shown by increased PLAI raw scores.

Regarding the different levels of inclusion, our statistically nonsignificant Treatment X Time interactions are consistent with findings of previous studies (Buysse & Bailey, 1993; Odom & McEvoy, 1988) in which preschool children with disabilities, on average, made comparable progress in special education-only and integrated special education classrooms (Cole et al., 1991; Jenkins, Odom, & Speltz, 1989; Jenkins, Speltz, & Odom, 1985) and in special education-only and mainstreamed classrooms (Fewell & Oelwein, 1990; Rule et al., 1987). On the other hand, our analysis of effect sizes and within-group pre- to posttest gains suggest that only one treatment, integrated special education, consistently produced growth that significantly exceeded normal development.

Cole et al. (1991) found ATIs between child characteristics and type of placement; specifically, higher performing children made relatively greater gains in integrated special education classrooms, and lower performing children made relatively greater gains in special education-only classrooms. In the present study, multiple regression analyses testing the same treatment conditions (i.e., integrated special education and special education-only) and child characteristics (i.e., pretest McCarthy GCI and PLAI scores) replicated the pattern of ATIs reported in our earlier study.

Were one to consider only the results of the integrated special education and special education-only classroom compositions in both the present study and that of Cole et al. (1991), it would be tempting to suppose that further increasing the ratio of typically developing children to children with disabilities would further improve the development of higher performing children with disabilities. The addition of the mainstream treatment in the present study permits an examination of this hypothesis. Mainstreamed classrooms had more than twice as many typically developing children than integrated special education classrooms (proportionally 64% were typically developing in the mainstream treatment vs. 21% in the integrated special education treatment). Rather than benefiting from the larger numbers of typically developing classmates, higher performing special education children in the mainstreamed classrooms showed less growth than their counterparts in the integrated special education classrooms. Higher performing children in mainstream classrooms performed no better than those in special education-only classrooms. Thus, increasing the proportion of typically developing children does not in itself appear to yield better outcomes for higher performing children.

These results indicate that the mix of typically developing to special education children provided by integrated special education classrooms (3 children who are typically developing and 11 children with disabilities) results in improved outcomes for higher performing children relative to outcomes produced by either higher (mainstreamed) or lower (special education-only) ratios of typically developing classmates. The nonlinear relation between child ratios and developmental outcome is not easily interpreted, but it appears that introducing a high ratio of typically developing children changes classroom dynamics in ways that do not favor higher functioning special education children. Perhaps the academic demands in mainstreamed classes exceed the abilities of the special education participants, or perhaps the ratio of typically developing children in mainstream classes affects child interaction patterns such that special education and typically developing children keep more to themselves.

It is also difficult to understand why children with lower language and cognitive skills achieved more in mainstreamed and special education-only treatments (extreme opposites), relative to their achievement in the integrated special education treatment. Perhaps these different student mixes induce teachers to establish instructional groupings that are more or less beneficial to low performing students. Alternatively, the different mixes may induce teachers to deliver instruction at levels which are more or less accessible to lower functioning students.

IMPLICATIONS

Our earlier findings (Cole et al., 1991) that lower functioning children benefited more from special education-only than from integrated special education raises a potential dilemma for practitioners, especially those who arc committed to inclusion on philosophical, legal, and social grounds. The results of the present study may help resolve this dilemma. Thc finding that lower performing children advanced comparably in mainstreamed and special education-only settings, combined with the finding that higher performing children benefited more from integrated settings suggest that both higher and lower performing groups can profit from some level of inclusion. However, the optimal degree of inclusion differs for these two groups; relatively higher functioning preschoolers with disabilities did not derive as much benefit from a full inclusion (i.e., mainstreamed) model as they did from an integrated model. In contrast, lower functioning preschoolers benefited as much from a full inclusion model as they did from a special education-only placement. These findings challenge the idea that one type of placement (full inclusion) is best for all children. By limiting placement for special education children to mainstreamed classrooms only, some children may experience a less than optimal learning environment. As Bricker (1995) reminds us, the needs of the child should not be lost in a movement to advocate one type of placement over all other considerations.

Transportation costs, personnel, geography, and population density affect school districts' ability to maintain a full continuum of services for preschool children with disabilities. In such cases, decisions on which program(s) to offer become important. Our results suggest that providing integrated and special education-only programs, or integrated and mainstream programs is most likely to meet the needs of preschool children who fall into the moderate range of disabilities.

Two aspects of this research strengthen confidence about the findings. First, our findings replicate and extend our earlier results which showed that children vary in their response to the same educational environments. Second, the research classrooms in this study employed a broad range of curricular approaches, one emphasizing cognitive strategies, another emphasizing academic learning, another direct language training, and another child-initiated language techniques. Thus, our findings would appear to have applicability across a variety of curricular approaches.

Still, caution is warranted in applying these results. The reliability of our findings would have been enhanced by adding growth measures between the pre- and posttest points. Moreover, the interaction between aptitude and treatment accounts for a relatively small proportion of the variance on cognitive and language measures.

REFERENCES

Becker, W. C. (1977). Teaching reading and language to the disadvantaged--What we have learned from field research. Harvard Educational Review, 47, 518-543.

Becker, W. C., Engelmann, S., & Thomas, D. R. (1975). Teaching 2: Cognitive learning and instruction. Chicago: Science Research Associates.(*)

Blank, M., Rose, S. A., & Berlin, L. J. (1978). Preschool language assessment instrument: The language of learning in practice. Orlando, FL: Grune & Stratton.(*)

Bricker, D. (1995). The challenge of inclusion. Journal of Early Intervention, 19(3) 179-194.

Buysse, V., & Bailey, D. B. (1993). Behavioral and developmental outcomes in young children with disabilities in integrated and special education-only settings: A review of comparative studies, The Journal of Special Education, 26(4), 434-461.

Cole, K. N., Dale, P. S., Mills, P. E., & Jenkins, J. R. (1993). Interaction between early intervention curricula and student characteristics. Exceptional Children, 60, 17-28.

Cole, K. N., Mills, P. E., Dale, P. S., & Jenkins, J. R. (1991). Effects of preschool integration for children with disabilities. Exceptional Children, 58, 36-45.

Cole, K. N., Mills, P. E., Dale, P. S., & Jenkins, J. R. (1996). Preschool language facilitation methods and child characteristics. Journal of Early Intervention, 20(2), 113-131.

Drummond, W. T. (1989). Enterprise method. Unpublished manuscript.

Fewell, R. R., & Oelwein, P. L. (1990). The relationship between time in integrated environments and developmental gains in young children with special needs. Topics in Early Childhood Special Education, 10(2), 104-116.

Fuchs, D., & Fuchs, L., S. (1994). Inclusive schools movement and the radicalization of special education reform. Exceptional Children, 60, 294-309.

Glass, G. V., McGaw, B., & Smith, M. L. (1981). Meta-analysis in social research. Beverly Hills, CA: Sage.(*)

Guralnick, M. J. (1990). Major accomplishments and future directions in early childhood mainstreaming. topics in Early Childhood Special Education, 10(2), 1-17.

Jenkins, J. R., Odom, S. L. & Speltz, M. L. (1989). Effects of social integration on preschool children with handicaps, Exceptional Children, 55, 420-428.

Jenkins, J. R., Speltz, M. L., & Odom, S. L. (1985). Integrating normal and handicapped preschoolers: Effects on child development and social interaction. Exceptional Children, 52, 7-17.

Lamorey, S., & Bricker, D. (1993). Integrated programs: Effects on young children and their parents. In C. Peck, S. Odom, & D. Bricker (Eds.), Integrating young children with disabilities into community programs: Ecological perspectives on research and implementation, (pp. 39-64). Baltimore: Paul H. Brookes.(*)

MacMillan, D. L., Gresham, F. M., & Forness, S. R. (1996). Full inclusion: An empirical perspective. Behavioral Disorders, 22 (2), 145-159.

McCarthy, D. (1972). McCarthy Scales of Children) Abilities. New York: The Psychological Corporation.(*)

Odom, S. L., & McEvoy, M. (1988). Integration of young children with handicaps and normally developing children. In S. L. Odom & M. B. Karnes (Eds.), Early intervention for infants and children with handicaps: An empirical base (pp. 241-267). Baltimore: Paul H. Brookes.(*)

Osborn J., Sherwood, D., & Cole, K. (1991). Mediated Learning Curriculum. Unpublished manuscript.

Peck, C. A., (1995). Some further reflections on the difficulties and dilemmas of inclusion. Journal of Early Intervention, 19(3), 197-199.

Pedhazur, E. J. (1982). Multiple regression in behavioral research (2nd ed.). New York: Holt, Rinehart, & Winston.(*)

Rule, S., Stowitschek, J. J., Innocenti, M., Striefel, S., Killoran, J., Swezey, K., & Boswell, C. (1987). The social integration program: An analysis of the effects of mainstreaming handicapped children into day care centers. Education and Treatment of Children, 10(2), 175-192.

Salisbury, C. L., & Vincent, L. J. (1990). Criterion of the next environment and best practices: Mainstreaming and integration 10 years later. Topics in Early Childhood Special Education, 10(2), 78-89.

Sattler, J. M. (1988). Assessment of children (3rd ed.). San Diego, CA: Author.(*)

Strain, P. S. (1995). The challenge of inclusion: Points well-taken and related challenges. Journal of Early Intervention, 19(3), 195-196.

Waryas, C., & Stremel-Campbell, K. (1984). Communication training program. New York: Teaching Resources.(*)

Zigmond, N., Jenkins, J., Fuchs, D., Deno, S., & Fuchs, L. S. (1995). When students fail to achieve satisfactorily: A reply to McClesky and Waldron. Phi Delta Kappan, 77(4), 303-306.

Zigmond, N., Jenkins, J., Fuchs, L. S., Deno, S., Fuchs, D., Baker, J. N., Jenkins, L., & Couthino, M. (1995). Special education in restructured schools: Findings from three multi-year studies. Phi Delta Kappan, 76(7), 531-540.

BooksNow

(*) To order books referenced in this journal please call 24 hrs/365 days: (800) BOOKS-NOW (266-5766) or [phone], or visit them on the Web at http://www.BooksNow.com/Exceptional Children.htm. Use Visa, M/C, or AMEX or send check or money order + $4.95 S&H ($2.50 each add'l item) to: BooksNow, 448 E. 6400 South, Suite 125, Salt Lake City, UT 84107.

PAULETTE E. MILLS, Assistant Professor, Department of Human Development, Washington State University, Pullman. KEVIN N. COLE, Senior Researcher, Washington Research Institute; JOSEPH R. JENKINS, Professor, Experimental Education Unit; PHILIP S. DALE, Professor, Department of Psychology, University of Washington, Seattle.

This research was supported by Grants H024A80030 and H024V00002 from the U.S. Department of Education to the University of Washington. Points of view or opinions stated in this report do not necessarily represent official agency positions.

Correspondence concerning this article should be addressed to Paulette E. Mills, Department of Human Development, 311 Hulbert Hall, Washington State University, P. O. Box 646236, Pullman, Washington 99164-6236. Electronic mail may be sent via Internet to [e-mail].

Manuscript received March 1997; revision accepted March 1998.