The Individuals with Disabilities Education Act (IDEA), 20 U.S.C. §§ 1400-1485 (formerly the "Education for All Handicapped Children Act"), provides that states receiving funding under the Act must ensure that children with disabilities are educated in regular classrooms with nondisabled children "to the maximum extent appropriate." 20 U.S.C. § 1412(5)(B). Plaintiff-appellee Rafael Oberti is an eight year old child with Down's syndrome who was removed from the regular classroom by defendant-appellant Clementon School District Board of Education (the "School District") and placed in a segregated special education class. In this appeal, we are asked by the School District to review the district court's decision in favor of Rafael and his co-plaintiff parents Carlos and Jeanne Oberti concerning Rafael's right under IDEA to be educated in a regular classroom with nondisabled classmates. This court has not previously had occasion to interpret or apply the "mainstreaming" requirement of IDEA.1

We construe IDEA's mainstreaming requirement to prohibit a school from placing a child with disabilities outside of a regular classroom if educating the child in the regular classroom, with supplementary aids and support services, can be achieved satisfactorily. In addition, if placement outside of a regular classroom is necessary for the child to receive educational benefit, the school may still be violating IDEA if it has not made sufficient efforts to include the child in school programs with nondisabled children whenever possible. We also hold that the school bears the burden of proving compliance with the mainstreaming requirement of IDEA, regardless of which party (the child and parents or the school) brought the claim under IDEA before the district court.

Although our interpretation of IDEA's mainstreaming requirement differs somewhat from that of the district court, we will affirm the decision of the district court that the School District has failed to comply with IDEA. More precisely, we will affirm the district court's order that the School District design an appropriate education plan for Rafael Oberti in accordance with IDEA, and we will remand for further proceedings consistent with this opinion. We do not reach the question, decided by the district court in favor of Rafael and his parents Carlos and Jeanne Oberti, whether § 504 of the Rehabilitation Act also supports relief, since, in view of our decision under IDEA, resolution of that issue is not necessary to the result.

Rafael is an eight year old child with Down's syndrome, a genetic defect that severely impairs his intellectual functioning and his ability to communicate. Now and throughout the period in question, Rafael and his parents have lived within the Clementon School District, in southern New Jersey. Prior to his entry into kindergarten, Rafael was evaluated in accordance with federal and state law by the School District's Child Study Team (the "Team"). See 20 U.S.C. § 1412(2)(C); N.J.A.C. 6:28-3.1--6:28-3.4.2 Based on its evaluation, the Team recommended to Rafael's parents that he be placed in a segregated special education class located in another school district for the 1989-90 school year. The Obertis visited a number of special classes recommended by the School District and found them all unacceptable. Thereafter the Obertis and the School District came to an agreement that Rafael would attend a "developmental" kindergarten class (for children not fully ready for kindergarten) at the Clementon Elementary School (Rafael's neighborhood school) in the mornings, and a special education class in another school district in the afternoons.

The Individualized Education Plan (IEP) developed by the School District for Rafael for the 1989-90 school year, see 20 U.S.C. §§ 1401(a)(20), 1414(a)(5); N.J.A.C. 6:28-3.6; infra n. 16, assigned all of Rafael's academic goals to the afternoon special education class. In contrast, the only goals for Rafael in the morning kindergarten class were to observe, model and socialize with nondisabled children.

While Rafael's progress reports for the developmental kindergarten class show that he made academic and social progress in that class during the year, Rafael experienced a number of serious behavioral problems there, including repeated toileting accidents, temper tantrums, crawling and hiding under furniture, and touching, hitting and spitting on other children. On several occasions Rafael struck at and hit the teacher and the teacher's aide.

These problems disrupted the class and frustrated the teacher, who consulted the school psychologist and other members of the Child Study Team to discuss possible approaches to managing Rafael's behavior problems. The teacher made some attempts to modify the curriculum for Rafael, but Rafael's IEP provided no plan for addressing Rafael's behavior problems. Neither did the IEP provide for special education consultation for the kindergarten teacher, or for communication between the kindergarten teacher and the special education teacher. In March of 1990, the School District finally obtained the assistance of an additional aide, which had been requested by the parents much earlier in the school year, but the presence of the extra aide in the kindergarten class did little to resolve the behavior problems. According to Rafael's progress reports for the afternoon special education class, and as the district court found, Rafael did not experience similar behavior problems in that class.

At the end of the 1989-90 school year, the Child Study Team proposed to place Rafael for the following year in a segregated special education class for children classified as "educable mentally retarded." Since no such special education class existed within the Clementon School District, Rafael would have to travel to a different district. The Team's decision was based both on the behavioral problems Rafael experienced during the 1989-90 school year in the developmental kindergarten class and on the Team's belief that Rafael's disabilities precluded him from benefiting from education in a regular classroom at that time.

The Obertis objected to a segregated placement and requested that Rafael be placed in the regular kindergarten class in the Clementon Elementary School. The School District refused, and the Obertis sought relief by filing a request for a due process hearing.3 The parties then agreed to mediate their dispute, pursuant to New Jersey regulations, as an alternative to a due process hearing. See N.J.A.C. 6:28-2.6. Through mediation, the Obertis and the School District came to an agreement that for the 1990-91 school year Rafael would attend a special education class for students labeled "multiply handicapped" in a public elementary school in the Winslow Township School District ("Winslow"), approximately 45 minutes by bus from Rafael's home. As part of the agreement, the School District promised to explore mainstreaming possibilities at the Winslow school and to consider a future placement for Rafael in a regular classroom in the Clementon Elementary School.4

The special education class in Winslow that Rafael attended during the 1990-91 school year was taught by an instructor and an instructional aide and included nine children. Although Rafael initially exhibited some of the same behavioral problems he had experienced in the Clementon kindergarten class, his behavior gradually improved: he became toilet trained and his disruptiveness abated. Rafael also made academic progress. However, by December of 1990, Rafael's parents found that the School District was making no plans to mainstream Rafael. The Obertis also learned that Rafael had no meaningful contact with nondisabled students at the Winslow school.5

In January of 1991, the Obertis brought another due process complaint, renewing their request under IDEA that Rafael be placed in a regular class in his neighborhood elementary school. A three-day due process hearing was held in February of 1991 before an Administrative Law Judge (ALJ) of the New Jersey Office of Administrative Law. See N.J.A.C. 6:28-2.7(e)4.iv; supra n. 3. On March 15, 1991, the ALJ affirmed the School District's decision that the segregated special education class in Winslow was the "least restrictive environment" for Rafael.6 Based on the testimony of Rafael's kindergarten teacher and other witnesses for the School District who described Rafael's disruptive behavior in the developmental kindergarten class, the ALJ found that Rafael's behavior problems in that class were extensive and that he had achieved no meaningful educational benefit in the class.7 The ALJ concluded that Rafael was not ready for mainstreaming.8

In reaching this conclusion, the ALJ discounted the testimony of the Obertis' two expert witnesses. Dr. Gail McGregor, a professor of education at Temple University and an expert in the education of children with disabilities, testified that Rafael could be educated satisfactorily in a regular class at the Clementon Elementary School with supplementary aids and services, and that Rafael would learn important skills in a regular classroom that could not be learned in a segregated setting.9 The ALJ disregarded Dr. McGregor's testimony because, unlike the School District's witnesses, she did not have daily experience with Rafael. Likewise, the ALJ discounted the testimony of the Obertis' other expert witness, Thomas Nolan, a teacher and special education specialist who had taught a child with Down's syndrome in a regular classroom, because he too had not had day-to-day experience with Rafael.10 The ALJ thus concluded that the Winslow placement was in compliance with IDEA.

C. The proceedings before the district court

Seeking independent review of the ALJ's decision pursuant to 20 U.S.C. § 1415(e)(2), the Obertis filed this civil action in the United States District Court for the District of New Jersey. In addition to the IDEA claim, the Obertis pleaded a claim of unlawful discrimination under § 504 of the Rehabilitation Act of 1973, 29 U.S.C. § 794. Soon thereafter, the court denied both parties' motions for summary judgment, finding "genuine issues of material fact ... about the feasibility of including Rafael in a regular classroom setting now." Oberti v. Board of Educ. of Clementon School Dist., 789 F. Supp. 1322, 1336 (D.N.J.1992) (Oberti I ).

In May of 1992, the district court held a three-day bench trial, receiving new evidence from both parties to supplement the state agency record. See 20 U.S.C. § 1415(e)(2).11 The Obertis presented the testimony of two additional experts who had not testified in the administrative proceedings: Dr. Lou Brown, a professor of special education at the University of Wisconsin, and Amy Goldman, an expert in communication with children with developmental disabilities.

Dr. Brown, who over the past twenty years has been a consultant to hundreds of school districts throughout the country regarding the education of severely disabled children, interviewed and evaluated Rafael on two occasions, and reviewed Rafael's educational records, as well as a set of videotapes showing Rafael at age seven working with his mother, being taught by a language professional, and participating in a Sunday school class with nondisabled children. Dr. Brown testified that he saw no reason why Rafael could not be educated at that time in a regular classroom with appropriate supplementary aids and services. He told the court that if such aids and services were provided, he had no reason to believe that Rafael would be disruptive at that time (more than two years after the experience in the Clementon kindergarten class). He also stated that integrating Rafael in a regular class at his local school would enable Rafael to develop social relationships with nondisabled students and to learn by imitating appropriate role models, important benefits which could not be realized in a segregated, special education setting.

Dr. Brown outlined a number of commonly applied strategies which could be used, in combination, by the School District to integrate Rafael in a regular classroom, including: (1) modifying some of the curriculum to accommodate Rafael's different level of ability; (2) modifying only Rafael's program so that he would perform a similar activity or exercise to that performed by the whole class, but at a level appropriate to his ability; (3) "parallel instruction," i.e., having Rafael work separately within the classroom on an activity beneficial to him while the rest of the class worked on an activity that Rafael could not benefit from; and (4) removing Rafael from the classroom to receive some special instruction or services in a resource room, completely apart from the class. Dr. Brown explained that with proper training, a regular teacher would be able to apply these techniques and that, in spite of Rafael's severe intellectual disability, a regular teacher with proper training would be able to communicate effectively with Rafael. Dr. Brown also testified that many of the special educational techniques applied in the segregated Winslow class could be provided for Rafael within a regular classroom.

Based on her evaluation of Rafael and her expertise in developing communication skills for disabled children, Amy Goldman testified that the speech and language therapy Rafael needs could be most effectively provided within a regular classroom; otherwise, she explained, a child with Rafael's disabilities would have great difficulty importing the language skills taught in a separate speech therapy session into the regular class environment, where those skills are most needed. She testified that language and speech therapy could easily be provided by a therapist inside the regular class during ongoing instruction if the therapist were able to collaborate ahead of time with the instructor regarding the upcoming lesson plans.

In addition, Dr. McGregor reaffirmed her prior opinion in the administrative proceedings that placement in a regular classroom was not only feasible but preferable for Rafael, see supra n. 9. Further, she testified that, given the resources and expertise available to public schools in New Jersey, the School District should be able to design an inclusive program for Rafael with assistance from professionals who have experience integrating children with disabilities in regular classes.

The Obertis also offered the videotape evidence that had been reviewed by Dr. Brown, the testimony of Jeanne Oberti,12 and the testimony of Joanne McKeon, the mother of a nine year old child with Down's syndrome who had been successfully mainstreamed in a regular classroom.

To counter the Obertis' experts, the School District offered Dr. Stanley Urban, a professor of special education at Glassboro State College. After observing Rafael in a special class for perceptually impaired children at the St. Luke's School (a private school that Rafael attended for two months in the fall of 1991), observing Rafael for two hours in his home, reviewing the programs available at the Clementon Elementary School, reviewing Rafael's education records, and reviewing the written evaluations of the Obertis' experts, Dr. Urban testified that in his opinion Rafael could not be educated satisfactorily in a regular classroom, and that the special education program at Winslow was appropriate for Rafael.13

More specifically, Dr. Urban testified that Rafael's behavior problems could not be managed in a regular class, that a regular teacher would not be able to communicate with a child of Rafael's ability level, and that it would be difficult if not impossible to adapt a first grade-level curriculum to accommodate Rafael without adversely affecting the education of the other children in the class. Dr. Urban, however, also stated that if Rafael did not have serious behavior problems, integration in a regular classroom might be feasible.

The School District presented several additional witnesses, including the teacher and teacher's aide of a non-academic summer school class for elementary school children which Rafael attended in the summer of 1991, and the teacher of the St. Luke's class, which Rafael attended for two months in the fall of 1991. These witnesses recounted examples of Rafael's disruptive behavior, including pushing and hitting other children, disobeying and running away from the instructors, and throwing books.

In August of 1992, after reviewing all of this new evidence along with the evidence that had been adduced at the administrative proceedings, the district court issued its decision, finding that the School District had failed to establish by a preponderance of the evidence that Rafael could not at that time be educated in a regular classroom with supplementary aids and services. The court therefore concluded that the School District had violated IDEA. Oberti v. Board of Educ. of Clementon School Dist., 801 F. Supp. 1392 (D.N.J.1992) (Oberti II ).

In particular, the court was persuaded by the Obertis' experts that many of the special education techniques used in the Winslow class could be implemented in a regular classroom. Id. at 1397. The court also found that the School District did not make reasonable efforts to include Rafael in a regular classroom with supplementary aids and services (e.g., an itinerant teacher trained in aiding students with mental retardation, a behavior management program, modification of the regular curriculum to accommodate Rafael, and special education training and consultation for the regular teacher); that Rafael's behavior problems during the 1989-90 school year in the developmental kindergarten class were largely the result of the School District's failure to provide adequate supplementary aids and services; and that the record did not support the School District's contention that Rafael would present similar behavior problems at that time (more than two years after the kindergarten class) if included in a regular classroom setting with adequate aids and services. Id. at 1397, 1403. The court declined to defer to the findings of the ALJ because it found that "they were largely and improperly based upon Rafael's behavior problems in the developmental kindergarten as well as upon his intellectual limitations, without proper consideration of the inadequate level of supplementary aids and services provided by the School District." Id. at 1404.

In addition to finding a violation of IDEA, the court concluded that by refusing to include Rafael in a regular classroom, the School District was discriminating against Rafael in violation of § 504 of the Rehabilitation Act. Id. at 1406-07. Accordingly, the court ordered the School District "to develop an inclusive plan for Rafael Oberti for the 1992-93 school year consistent with the requirements" of IDEA and § 504 of the Rehabilitation Act. This appeal followed. We have jurisdiction under 28 U.S.C. § 1291. The order of the district court has been stayed pending appeal.14 See 20 U.S.C. § 1415(e)(3).

The Education for All Handicapped Children Act (IDEA's predecessor statute) was enacted in 1975 in response to a Congressional finding that "more than half of the children with disabilities in the United States do not receive appropriate educational services." 20 U.S.C. § 1400(b)(3); see also S.Rep. No. 168, 94th Cong., 1st Sess. 8 (1975), reprinted in 1975 U.S.C.C.A.N. 1425, 1432. The Act provides federal funds to participating states for the education of children with disabilities.15 As a condition of receiving these funds, states must have "in effect a policy that assures all children with disabilities the right to a free appropriate public education." 20 U.S.C. § 1412(1).

In Board of Educ. v. Rowley, 458 U.S. 176, 188-89, 102 S. Ct. 3034, 3042, 73 L. Ed. 2d 690 (1982), the Supreme Court held that a "free appropriate public education" under the Act "consists of educational instruction specially designed to meet the unique needs of the handicapped child, supported by such services as are necessary to permit the child 'to benefit' from the instruction." This court in turn interpreted Rowley to require the state to offer children with disabilities individualized education programs that provide more than a trivial or de minimis educational benefit. Polk v. Central Susquehanna Intermediate Unit 16, 853 F.2d 171, 180-85 (3d Cir.1988), cert. denied, 488 U.S. 1030, 109 S. Ct. 838, 102 L. Ed. 2d 970 (1989).16

In addition to the free appropriate education requirement, IDEA provides that states must establish

procedures to assure that, to the maximum extent appropriate, children with disabilities ... are educated with children who are not disabled, and that special classes, separate schooling, or other removal of children with disabilities from the regular educational environment occurs only when the nature and severity of the disability is such that education in regular classes with the use of supplementary aids and services cannot be achieved satisfactorily....

20 U.S.C. § 1412(5)(B).17 As numerous courts have recognized, this provision sets forth a "strong congressional preference" for integrating children with disabilities in regular classrooms. See, e.g., Devries v. Fairfax County School Bd., 882 F.2d 876, 878 (4th Cir.1989); Daniel R.R. v. State Bd. of Educ., 874 F.2d 1036, 1044 (5th Cir.1989); A.W. v. Northwest R-1 School Dist., 813 F.2d 158, 162 (8th Cir.1987); Roncker v. Walter, 700 F.2d 1058, 1063 (6th Cir.), cert. denied, 464 U.S. 864, 104 S. Ct. 196, 78 L. Ed. 2d 171 (1983); Board of Educ. Sacramento City Unified School Dist. v. Holland, 786 F. Supp. 874, 878 (E.D.Cal.1992).

One of our principal tasks in this case is to provide standards for determining when a school's decision to remove a child with disabilities from the regular classroom and to place the child in a segregated environment violates IDEA's presumption in favor of mainstreaming. This issue is particularly difficult in light of the apparent tension within the Act between the strong preference for mainstreaming, 20 U.S.C. § 1412(5)(B), and the requirement that schools provide individualized programs tailored to the specific needs of each disabled child, 20 U.S.C. §§ 1401, 1414(a)(5). See Daniel R.R., 874 F.2d at 1044; Greer v. Rome City School Dist., 950 F.2d 688, 695 (11th Cir.1991).18

The key to resolving this tension appears to lie in the school's proper use of "supplementary aids and services," 20 U.S.C. § 1412(5)(B), which may enable the school to educate a child with disabilities for a majority of the time within a regular classroom, while at the same time addressing that child's unique educational needs. We recognize, however, that "[r]egular classes ... will not provide an education that accounts for each child's particular needs in every case." Daniel R.R., 874 F.2d at 1044; see also Devries, 882 F.2d at 878-80 (holding that 17 year old autistic student could not benefit from "monitoring" regular high school academic classes and was appropriately placed at county vocational center).

We also recognize that "[i]n assuring that the requirements of the Act have been met, courts must be careful to avoid imposing their view of preferable educational methods upon the States." Rowley, 458 U.S. at 207, 102 S.Ct. at 3051. We are mindful that the Act leaves questions of educational policy to state and local officials. Id. On the other hand, as the Supreme Court recognized in Rowley, the Act specifically "requires participating States to educate handicapped children with nonhandicapped children whenever possible." Rowley, 458 U.S. at 202, 102 S.Ct. at 3049; see also Honig v. Doe, 484 U.S. 305, 311, 108 S. Ct. 592, 597, 98 L. Ed. 2d 686 (1988).19 It is our duty to enforce that statutory requirement. See Polk, 853 F.2d at 184 ("We do not read the Supreme Court's salutary warnings against interference with educational methodology as an invitation to abdicate our obligation to enforce the statutory provisions [of the Act].").20

In Daniel R.R., the Fifth Circuit derived from the language of 20 U.S.C. § 1412(5)(B) a two-part test for determining whether a school is in compliance with IDEA's mainstreaming requirement. First, the court must determine "whether education in the regular classroom, with the use of supplementary aids and services, can be achieved satisfactorily." 874 F.2d at 1048.21 Second, if the court finds that placement outside of a regular classroom is necessary for the child to benefit educationally, then the court must decide "whether the school has mainstreamed the child to the maximum extent appropriate," i.e., whether the school has made efforts to include the child in school programs with nondisabled children whenever possible. Id. We think this two-part test, which closely tracks the language of § 1412(5)(B), is faithful to IDEA's directive that children with disabilities be educated with nondisabled children "to the maximum extent appropriate," 20 U.S.C. § 1412(5)(B), and to the Act's requirement that schools provide individualized programs to account for each child's specific needs, 20 U.S.C. §§ 1401, 1414(a)(5). See Greer, 950 F.2d at 696 (adopting the Daniel R.R. test); Liscio v. Woodland Hills School Dist., 734 F. Supp. 689 (W.D.Pa.1989) (same).

The district court in this case adopted the somewhat different test set forth by the Sixth Circuit in Roncker v. Walter, 700 F.2d 1058 (6th Cir.1983), the first federal court of appeals case to interpret IDEA's mainstreaming requirement. See Oberti II, 801 F. Supp. at 1401. In Roncker, the court stated:

In a case where the segregated facility is considered superior [academically], the court should determine whether the services which make that placement superior could be feasibly provided in a non-segregated setting. If they can, the placement in the segregated school would be inappropriate under the Act.

700 F.2d at 1063; see also A.W. v. Northwest R-1 School Dist., 813 F.2d 158, 163 (8th Cir.1987) (adopting Roncker test). We believe, however, that the two-part Daniel R.R. test is the better standard because the Roncker test fails to make clear that even if placement in the regular classroom cannot be achieved satisfactorily for the major portion of a particular child's education program, the school is still required to include that child in school programs with nondisabled children (specific academic classes, other classes such as music and art, lunch, recess, etc.) whenever possible. We therefore adopt the two-part Daniel R.R. test rather than the standard espoused in Roncker.

In applying the first part of the Daniel R.R. test, i.e., whether the child can be educated satisfactorily in a regular classroom with supplementary aids and services, the court should consider several factors. First, the court should look at the steps that the school has taken to try to include the child in a regular classroom. See Greer, 950 F.2d at 696; Daniel R.R., 874 F.2d at 1048. As we have explained, the Act and its regulations require schools to provide supplementary aids and services to enable children with disabilities to learn whenever possible in a regular classroom. See 20 U.S.C. §§ 1401(a)(17), 1412(5)(B); 34 C.F.R. § 300.551(b)(2). The regulations specifically require school districts to provide "a continuum of placements ... to meet the needs of handicapped children." 34 C.F.R. § 300.551(a). The continuum must "[m]ake provision for supplementary services (such as resource room22 or itinerant instruction) to be provided in conjunction with regular class placement." 34 C.F.R. § 300.551(b).

Accordingly, the school "must consider the whole range of supplemental aids and services, including resource rooms and itinerant instruction," Greer, 950 F.2d at 696, speech and language therapy, special education training for the regular teacher, behavior modification programs, or any other available aids or services appropriate to the child's particular disabilities. The school must also make efforts to modify the regular education program to accommodate a disabled child. See 34 C.F.R. Part 300, App.C. Question 48. If the school has given no serious consideration to including the child in a regular class with such supplementary aids and services and to modifying the regular curriculum to accommodate the child, then it has most likely violated the Act's mainstreaming directive. "The Act does not permit states to make mere token gestures to accommodate handicapped students; its requirement for modifying and supplementing regular education is broad." Daniel R.R., 874 F.2d at 1048; see also Greer, 950 F.2d at 696.

A second factor courts should consider in determining whether a child with disabilities can be included in a regular classroom is the comparison between the educational benefits the child will receive in a regular classroom (with supplementary aids and services) and the benefits the child will receive in the segregated, special education classroom. The court will have to rely heavily in this regard on the testimony of educational experts. Nevertheless, in making this comparison the court must pay special attention to those unique benefits the child may obtain from integration in a regular classroom which cannot be achieved in a segregated environment, i.e., the development of social and communication skills from interaction with nondisabled peers. See Daniel R.R., 874 F.2d at 1049 ("a child may be able to absorb only a minimal amount of the regular education program, but may benefit enormously from the language models that his nonhandicapped peers provide"); Greer, 950 F.2d at 697 (language and role modeling from association with nondisabled peers are essential benefits of mainstreaming); Holland, 786 F. Supp. at 882 (benefits obtained by child with mental retardation as result of placement in a regular classroom include development of social and communications skills and generally improved self-esteem).23 As IDEA's mainstreaming directive makes clear, Congress understood that a fundamental value of the right to public education for children with disabilities is the right to associate with nondisabled peers.24

Thus, a determination that a child with disabilities might make greater academic progress in a segregated, special education class may not warrant excluding that child from a regular classroom environment. We emphasize that the Act does not require states to offer the same educational experience to a child with disabilities as is generally provided for nondisabled children. See Rowley, 458 U.S. at 189, 202, 102 S. Ct. at 3042, 3048-49. To the contrary, states must address the unique needs of a disabled child, recognizing that that child may benefit differently from education in the regular classroom than other students. See Daniel R.R., 874 F.2d at 1047. In short, the fact that a child with disabilities will learn differently from his or her education within a regular classroom does not justify exclusion from that environment.

A third factor the court should consider in determining whether a child with disabilities can be educated satisfactorily in a regular classroom is the possible negative effect the child's inclusion may have on the education of the other children in the regular classroom. While inclusion of children with disabilities in regular classrooms may benefit the class as a whole, see supra n. 24, a child with disabilities may be "so disruptive in a regular classroom that the education of other students is significantly impaired." 34 C.F.R. § 300.552 comment (citing 34 C.F.R. part 104--Appendix, Para. 24); see Greer, 950 F.2d at 697; Daniel R.R., 874 F.2d at 1048-49. Moreover, if a child is causing excessive disruption of the class, the child may not be benefiting educationally in that environment. Accordingly, if the child has behavioral problems, the court should consider the degree to which these problems may disrupt the class. In addition, the court should consider whether the child's disabilities will demand so much of the teacher's attention that the teacher will be required to ignore the other students. See Daniel R.R., 874 F.2d at 1049.

We emphasize, however, that in considering the possible negative effect of the child's presence on the other students, the court must keep in mind the school's obligation under the Act to provide supplementary aids and services to accommodate the child's disabilities. See Greer, 950 F.2d at 697. An adequate individualized program with such aids and services may prevent disruption that would otherwise occur. See id. With respect to the concerns of nondisabled children in the regular classroom, we note that the comment to 34 C.F.R. § 300.552 (citing 34 C.F.R. part 104--Appendix, Para. 24) reads: "[I]t should be stressed that, where a handicapped child is so disruptive in a regular classroom that the education of other students is significantly impaired, the needs of the handicapped child cannot be met in that environment. Therefore, regular placements would not be appropriate to his or her needs...." On the other hand, "a handicapped child who merely requires more teacher attention than most other children is not likely to be so disruptive as to significantly impair the education of other children." Greer, 950 F.2d at 697.

In sum, in determining whether a child with disabilities can be educated satisfactorily in a regular class with supplemental aids and services (the first prong of the two-part mainstreaming test we adopt today), the court should consider several factors, including: (1) whether the school district has made reasonable efforts to accommodate the child in a regular classroom; (2) the educational benefits available to the child in a regular class, with appropriate supplementary aids and services, as compared to the benefits provided in a special education class; and (3) the possible negative effects of the inclusion of the child on the education of the other students in the class.25

If, after considering these factors, the court determines that the school district was justified in removing the child from the regular classroom and providing education in a segregated, special education class, the court must consider the second prong of the mainstreaming test--whether the school has included the child in school programs with nondisabled children to the maximum extent appropriate. See Daniel R.R., 874 F.2d at 1048, 1050. IDEA and its regulations "do not contemplate an all-or-nothing educational system in which handicapped children attend either regular or special education." Id. at 1050. The regulations under IDEA require schools to provide a "continuum of alternative placements ... to meet the needs of handicapped children." 34 C.F.R. § 300.551(a). As the Fifth Circuit stated:

the school must take intermediate steps wherever appropriate, such as placing the child in regular education for some academic classes and in special education for others, mainstreaming the child for nonacademic classes only, or providing interaction with nonhandicapped children during lunch and recess. The appropriate mix will vary from child to child and, it may be hoped, from school year to school year as the child develops.

Daniel R.R., 874 F.2d at 1050. Thus, even if a child with disabilities cannot be educated satisfactorily in a regular classroom, that child must still be included in school programs with nondisabled students wherever possible.

III. BURDEN OF PROOF UNDER IDEA's MAINSTREAMING REQUIREMENT

Before we apply the two-part analysis discussed above to the facts in this case, we must address the School District's argument that the district court improperly placed the burden of proof under the Act on it. In the School District's view, while it may have had the initial burden at the state administrative level of justifying its educational placement, once the agency decided in its favor, the burden should have shifted to the parents who challenged the agency decision in the district court. Courts must place the burden on the party seeking to reverse the agency decision, the School District argues, in order to effectuate IDEA's requirement that "due weight shall be given to [the state administrative] proceedings," Rowley, 458 U.S. at 206, 102 S.Ct. at 3051 (interpreting 20 U.S.C. § 1415(e)). We disagree.

IDEA instructs district courts and state trial courts reviewing the decisions of state educational agencies to "receive the records of the administrative proceedings.... hear additional evidence at the request of a party, and, basing its decision on the preponderance of the evidence, ... grant such relief as the court determines is appropriate." 20 U.S.C. § 1415(e)(2). As construed by the Supreme Court in Rowley, § 1415(e)(2) requires courts to give "due weight" to the agency proceedings. Rowley, 458 U.S. at 206, 102 S.Ct. at 3051. However, neither Rowley nor the Act itself specifically addresses which party bears the burden of proof at the district court level, an issue which we believe is quite different from the district court's obligation to afford due weight to the administrative proceedings.

The School District points to several cases that hold, either directly or implicitly, that even if the school district bears the burden of proving compliance with IDEA in the state administrative proceedings, the burden of proof shifts to the party challenging the agency decision at the district court level. See Roland M. v. Concord School Comm., 910 F.2d 983, 991 (1st Cir.1990), cert. denied, --- U.S. ----, 111 S. Ct. 1122, 113 L. Ed. 2d 230 (1991); Kerkam v. McKenzie, 862 F.2d 884, 887 (D.C.Cir.1988); see also Briggs v. Bd. of Educ., 882 F.2d 688, 692 (2d Cir.1989) (placing burden on parents challenging state agency decision to prove that their child's educational needs could be met in a less segregated setting). We find these cases unpersuasive.

In reviewing the decision of a state agency under IDEA, the district court "must make an independent determination based on a preponderance of the evidence." Geis v. Bd. of Educ., 774 F.2d 575, 583 (3d Cir.1985). Given that the district court must independently review the evidence adduced at the administrative proceedings and can receive new evidence, we see no reason to shift the ultimate burden of proof to the party who happened to have lost before the state agency, especially since the loss at the administrative level may have been due to incomplete or insufficient evidence or to an incorrect application of the Act.

The purpose of the "due weight" obligation is to prevent the court from imposing its own view of preferable educational methods on the states. See Rowley, 458 U.S. at 207, 102 S.Ct. at 3051. Accordingly, the due weight is owed to the administrative proceedings, not to the party who happened to prevail in those proceedings. Moreover, the amount of deference to be afforded the administrative proceedings "is an issue left to the discretion of the district court.... [T]he district court must consider the administrative findings of fact, but is free to accept or reject them." Jefferson County Bd. of Educ. v. Breen, 853 F.2d 853, 857 (11th Cir.1988). The district court can give due weight to the agency proceedings (i.e., refrain from imposing its own notions of educational policy on the states), while the ultimate burden of proof remains on the school.

Underlying the Act is "an abiding concern for the welfare of handicapped children and their parents." Lascari, 560 A.2d at 1188; see 20 U.S.C. § 1400(c).26 Requiring parents to prove at the district court level that the school has failed to comply with the Act would undermine the Act's express purpose "to assure that the rights of children with disabilities and their parents are protected," 20 U.S.C. § 1400(c), and would diminish the effect of the provision that enables parents and guardians to obtain judicial enforcement of the Act's substantive and procedural requirements, see 20 U.S.C. § 1415(e). In practical terms, the school has an advantage when a dispute arises under the Act: the school has better access to the relevant information, greater control over the potentially more persuasive witnesses (those who have been directly involved with the child's education), and greater overall educational expertise than the parents. See Lascari, 560 A.2d at 1188 (placing burden of proof on school is "consistent with the proposition that the burdens of persuasion and production should be placed on the party better able to meet those burdens."); Engel, Law, Culture, and Children with Disabilities, 1991 Duke L.J. at 187-94 (arguing that parents are generally at a disadvantage vis-a-vis the school when disputes arise under IDEA because parents generally lack specialized training and because their views are often treated as "inherently suspect" due to the attachment to their child).

In light of the statutory purpose of IDEA and these practical considerations, we believe that when IDEA's mainstreaming requirement is specifically at issue, it is appropriate to place the burden of proving compliance with IDEA on the school. Indeed, the Act's strong presumption in favor of mainstreaming, 20 U.S.C. § 1422(5)(B), would be turned on its head if parents had to prove that their child was worthy of being included, rather than the school district having to justify a decision to exclude the child from the regular classroom. See supra Part II. We therefore hold that the district court correctly placed the burden on the School District to prove that the segregated placement proposed for Rafael was in compliance with the mainstreaming requirement of IDEA.

IV. DID THE SCHOOL DISTRICT COMPLY WITH IDEA?

We now reach the dispositive question in this case: whether the district court erred in holding that the School District failed to comply with IDEA's mainstreaming requirement. Initially, applying the first part of the two-part test set forth above, supra Part II, we consider whether the School District has met its burden of proving that Rafael could not be educated satisfactorily in a regular classroom with supplementary aids and services.

Our deferential standard of review over the district court's fact finding is essential to our decision here. We accept the district court's findings of fact unless they are clearly erroneous. See David D. v. Dartmouth School Comm., 775 F.2d 411, 415 (1st Cir.1985), cert. denied, 475 U.S. 1140, 106 S. Ct. 1790, 90 L. Ed. 2d 336 (1986); Fed.R.Civ.P. 52(a). A finding of fact is clearly erroneous when, after reviewing the evidence, the court of appeals is "left with a definite and firm conviction that a mistake has been committed." Anderson v. Bessemer City, 470 U.S. 564, 573, 105 S. Ct. 1504, 1511, 84 L. Ed. 2d 518 (1985). Thus, even if we might have come to different factual conclusions based on this record, we defer to the findings of the district court unless we are convinced that the record cannot support those findings.

We also consider, in reviewing the district court's fact findings in an IDEA case, whether the court has abused its discretion in failing to afford "due weight" to the agency proceedings. See 20 U.S.C. § 1415(e)(2); Rowley, 458 U.S. at 206, 102 S.Ct. at 3051; Breen, 853 F.2d at 857. As discussed above, see supra Part III, the district court makes fact findings in an IDEA case not only on the administrative record, but also on any new evidence presented by the parties. The district court must give due weight to the administrative proceedings (so as not to impose the court's own view of preferable educational policy on the states), see Rowley, 458 U.S. at 206, 102 S.Ct. at 3051, but the court is free to accept or reject the agency findings depending on whether those findings are supported by the new, expanded record and whether they are consistent with the requirements of the Act. Accordingly, we review the district court's decisions regarding whether to adopt the agency fact findings under the deferential, clearly erroneous standard. See Breen, 853 F.2d at 857.

In Part II of this opinion, we outlined three factors that should be considered by a court in determining whether a child with disabilities can be educated satisfactorily in a regular classroom (the first part of the Daniel R.R. test): (1) whether the school district has made reasonable efforts to accommodate the child in a regular classroom with supplementary aids and services; (2) a comparison of the educational benefits available in a regular class and the benefits provided in a special education class; and (3) the possible negative effects of inclusion on the other students in the class. We now consider each of these factors, looking to the relevant fact findings of the district court to determine whether those findings are clearly erroneous, and if not, whether they support the district court's ultimate legal conclusion that the School District violated the mainstreaming requirement of IDEA.

As to the first factor, the district court found that the School District made only negligible efforts to include Rafael in a regular classroom. Specifically, the court found that during the 1989-90 school year, the only period during which the School District mainstreamed Rafael in a regular classroom, the School District placed Rafael in the developmental kindergarten class "without a curriculum plan, without a behavior management plan, and without providing adequate special education support to the teacher." Oberti II, 801 F. Supp. at 1402; see also id. at 1396, 1398. Further, the court found that the School District has since refused to include Rafael in a regular classroom largely based on the behavioral problems experienced by Rafael in the kindergarten class during the 1989-90 school year. Id. at 1396, 1403. For the 1990-91 year, the court found that Rafael was placed in a segregated class with "no meaningful mainstreaming opportunities," id. at 1397, and that "[t]he School District's consideration of less restrictive alternatives for the 1990-91 school year was perfunctory." Id. at 1396.

There is very little evidence in the record that conflicts with these findings. The School District produced some evidence that the kindergarten teacher and the school psychologist attempted to modify the curriculum in that class and to come up with methods of controlling Rafael's behavior problems. See supra n. 7. However, the record reflects that the School District had access to information and expertise about specific methods and services to enable children with disabilities like Rafael to be included in a regular classroom, see supra at 1211, but that the School District did not provide such supplementary aids and services for Rafael in the kindergarten class.27

Rafael's IEP for the 1989-90 school year included no provisions for supplementary aids and services in the kindergarten class aside from stating that there will be "modification of regular class expectations" to reflect Rafael's disability. The only goal provided for the regular kindergarten teacher was to "facilitate Rafael's adjustment to the kindergarten classroom." After reviewing this IEP along with the rest of Rafael's education records, Dr. McGregor testified that no supplementary aids and services were provided for Rafael in the 1989-90 kindergarten class. See supra n. 9.

Moreover, there is no evidence in the record that the School District gave any serious consideration to including Rafael in a regular classroom with supplementary aids and services after the 1989-90 school year; and the School District does not appear to dispute this fact. Further, Nancy Leech, the Winslow speech therapist (and one of the School District's witnesses) admitted that Rafael had not been included in any school programs with nondisabled children at Winslow, apart from attending lunch and school assemblies.

In view of the foregoing, the district court's finding that the School District has not taken meaningful steps to try to include Rafael in a regular classroom with supplementary aids and services is not clearly erroneous. We also note that the district court did not fail to give "due weight" to the agency proceedings on this issue since the ALJ did not even consider whether the School District had made efforts to include Rafael in a regular classroom with supplementary aids and services, as is required by IDEA. See Greer, 950 F.2d at 698 (school district's determination that child with Down's syndrome would receive more benefit in a segregated special education class "is due no deference because school officials failed to consider what benefits she would receive from education in a regular classroom with appropriate supplemental aids and services.") (emphasis in original). Accordingly, the School District's failure to give adequate consideration to including Rafael in a regular classroom with supplementary aids and services supports the district court's legal conclusion that the School District violated IDEA.

As to the second factor--a comparison of the educational benefits of the segregated placement in Winslow with the benefits Rafael could obtain from placement in a regular classroom--the district court found that "[v]arious experts who testified on Rafael's behalf have convincingly refuted the School District's assertion that such services could not be delivered within the matrix of a regular education class without disrupting the class or converting it into a special education class." Oberti II, 801 F. Supp. at 1403 n. 17. The court also found that Rafael would benefit academically and socially from inclusion in a regular classroom. Id. at 1404. Moreover, the district court found, based on expert testimony, that "nondisabled children in the class will likewise benefit" from the inclusion of Rafael in a regular classroom. Id. at 1404; see also supra n. 24.

The School District points to some evidence in the record that conflicts with these findings. Specifically, Dr. Urban testified for the School District that, in his opinion, a regular teacher would not be able to communicate with Rafael and that a regular curriculum would have to be modified beyond recognition to accommodate Rafael. See supra Part I.C. However, the Obertis' experts, Drs. Brown and McGregor, described various commonly applied methods that could be used to educate Rafael in a regular classroom and testified that, although Rafael has severe intellectual disabilities, a regular teacher with appropriate training would be able to communicate effectively with Rafael. See supra Parts I.B & C. They testified that many of the special education techniques used in the segregated Winslow class could be imported successfully into a regular classroom and that the regular teacher could be trained to apply these techniques. Id. Further, the Obertis' experts testified at length that inclusion in a classroom with nondisabled students would benefit Rafael substantially. Id. In addition, Amy Goldman testified that speech and language therapy not only could be provided in a regular classroom, but would be more effective for Rafael in an integrated setting. See supra Part I.C.

In short, the parties' experts disagreed on the respective benefits of a segregated versus an integrated placement for Rafael, and the district court was in a better position than we are to evaluate their testimony. We therefore defer to that court's findings, which, at all events, are not clearly erroneous. We note also that the district court did not fail to give due weight to the agency proceedings on this factor since the court's findings were based largely on new expert testimony that was not before the ALJ. Additionally, we agree with the district court's legal conclusion that, although including Rafael in a regular classroom would require the School District to modify the curriculum, the need for such modification is "not a legitimate basis upon which to justify excluding a child" from the regular classroom unless the education of other students is significantly impaired. Oberti II, 801 F. Supp. at 1403; see also 34 C.F.R. Part 300, App.C Question 48 (school must set forth in the IEP any modifications of the regular education program necessary to accommodate a disabled child). Thus, a comparison of the educational benefits of a segregated versus an integrated placement for Rafael supports the district court's conclusion that the School District's selection of a segregated placement did not comply with IDEA.

As to the third factor--the potentially disruptive effect of Rafael's presence on the other children in a regular classroom--the record again contains conflicting evidence. The School District presented numerous witnesses before both the ALJ and the district court who testified to Rafael's extremely disruptive behavior in the 1989-90 kindergarten class and in several other teaching environments. See supra Parts I.B & C. In contrast, the Obertis' experts Drs. McGregor and Brown evaluated Rafael and testified that in their opinion he would not at that point in time (nearly two years after Rafael's experience in the kindergarten class) cause any significant disruption in a regular classroom if provided with adequate supplementary aids and services, such as the assistance of an itinerant instructor with special education training, special education training for the regular teacher, modification of some of the academic curriculum to accommodate Rafael's disabilities, parallel instruction to allow him to learn at his academic level, and use of a resource room.

After evaluating the evidence on both sides, the district court found that "[t]here is nothing in the record which would suggest that at this point in time Rafael would present similar behavior problems if provided with an adequate level of supplementary aids and related services within the matrix of a regular education class. In fact, the record supports the opposite conclusion." Oberti II, 801 F. Supp. at 1403 (emphasis added). The court found that the behavioral problems Rafael experienced during the 1989-90 school year in the developmental kindergarten class "were exacerbated and remained uncontained due to the inadequate level of services provided there," that Rafael's behavioral problems were diminished in settings where an adequate level of supplementary aids and services were provided, and that both the School District and the ALJ "improperly justified Rafael's exclusion from less restrictive placements in subsequent years based upon those behavior problems." Id.

Although the School District presented ample evidence of Rafael's disruptive behavior in the 1989-90 kindergarten class, the Obertis' evidence supports the district court's finding that Rafael would not have had such severe behavior problems had he been provided with adequate supplementary aids and services in that kindergarten class, and that Rafael (who at the time of the district court trial was two years older than when he attended the kindergarten class) would most likely not present such problems if he were included in a regular class at that time. We therefore conclude that the district court's findings on this issue are not clearly erroneous, and, accordingly, that consideration of the possible negative effects of Rafael's presence on the regular classroom environment does not support the School District's decision to exclude him from the regular classroom.

We also conclude that the district court did not abuse its discretion in deciding not to defer to the findings of the ALJ on the issue of whether Rafael would significantly disrupt a regular classroom. As the court noted, the ALJ's findings "were largely and improperly based upon Rafael's behavior problems in the developmental kindergarten, as well as upon his intellectual limitations, without proper consideration of the inadequate level of supplementary aids and services provided by the School District." Oberti II, 801 F. Supp. at 1404 (emphasis added).28

For all of these reasons, we agree with the district court's conclusion that the School District did not meet its burden of proving by a preponderance of the evidence that Rafael could not be educated satisfactorily in a regular classroom with supplementary aids and services. We will therefore affirm the district court's decision that the School District has violated the mainstreaming requirement of IDEA. Because we have come to this conclusion based on application of the first part of the Daniel R.R. two-part test, we need not apply the second part of the test (whether the child has been included in programs with nondisabled children whenever possible).29 We note, however, that in the event that the Child Study Team were to determine in designing an IEP for Rafael in the future that education in a regular classroom with supplementary aids and services could not be achieved satisfactorily at that time and therefore would not be required under IDEA, the Team would then have to satisfy the second part of the Daniel R.R. test, ensuring that Rafael is included in regular school programs with nondisabled students whenever possible.30

Finally, in affirming the district court's order that the School District develop a more inclusive program for Rafael in compliance with IDEA for the upcoming school year, we emphasize that neither this court nor the district court is mandating a specific IEP for Rafael.31 The development of Rafael's IEP, and the specific nature of his placement, is, of course, the job of the Child Study Team.

The order of the district court will be affirmed.