This is a civil rights class action attacking the policies and practices of the Georgia State Board of Education and a number of local school districts. The plaintiffs, thirty-five black schoolchildren, sued in the United States District Court for the Southern District of Georgia claiming that black students are assigned to regular classes and programs for the educable mentally retarded (EMR) in Georgia public schools in a discriminatory manner. The district court found in favor of the defendants on all counts and the plaintiffs appealed that judgment to this court.

This action was filed on June 8, 1982 by the Georgia State Conference of Branches of the NAACP, the Liberty County Branch of the NAACP, and forty-five individual schoolchildren on behalf of themselves and all other black students similarly situated in Georgia except those living in Fulton, DeKalb, Cobb, Gwinnett, Clayton, Muscogee, Bibb, Richmond and Chatham counties. Ten of the children withdrew as plaintiffs shortly after the filing of the complaint.

The named defendants were the State of Georgia, the State Superintendent of Schools, the State Board of Education (the State defendants) and the Americus, Bleckley, Burke, Coweta, Crisp, Evans, Jefferson, Lee, Liberty, Miller, Thomaston, Pelham and Vidalia school districts (the local defendants). The claims against the Americus, Bleckley, Jefferson, Thomaston and Pelham school systems were settled prior to the trial.

The plaintiffs basically assert two causes of action. First, they allege that the use of achievement grouping in Georgia public schools is intended to achieve or results in intraschool racial segregation. This, they say, is violative of the thirteenth and fourteenth amendments to the Constitution of the United States; Title VI of the Civil Rights Act of 1964, 42 U.S.C. Secs. 2000d et seq. (Title VI); and the Equal Educational Opportunities Act, 20 U.S.C. Secs. 1701 et seq. (EEOA). The second cause of action incorporates the charge that black children in Georgia are assigned to EMR programs in a discriminatory manner. To the extent this claim is based on racial discrimination it is brought pursuant to the thirteenth and fourteenth amendments, Title VI and the EEOA.1 The allegations of handicap discrimination are founded on a violation of section 504 of the Rehabilitation Act of 1973, 29 U.S.C. Sec. 794 (section 504). Relief is sought against the local defendants for the practices within their districts and the State defendants are sued in their supervisory capacities.2

On April 5, 1983, 99 F.R.D. 16 (D.C.Ga.), the district court found that all the individual plaintiffs, with the exception of one from the Vidalia school district and five Coweta County students, were proper class representatives. The court also held the State Conference of Branches of the NAACP and the Liberty County Branch of the NAACP were not proper class representatives. The court thereupon certified the named plaintiffs as proper class representatives of "all black children in the State of Georgia" for the purposes of the claims against the State defendants and "the black children in the respective school system[s]" for the causes of action against the local defendants. Sealed Order of April 5, 1983 at 53.

The case was tried without a jury from October 31, 1983 to December 20, 1983. On June 28, 1984, the district court entered an order finding no infringement of the constitution, Title VI or the EEOA. The court, however, held that the defendants did violate certain regulations promulgated under section 504 governing the identification, evaluation and placement of students in EMR programs and scheduled a hearing for the appropriate relief.

On July 31, 1984, the defendants filed a motion to alter or amend the judgment, citing Smith v. Robinson, 468 U.S. ----, 104 S. Ct. 3457, 82 L. Ed. 2d 746 (1984), a case which had been decided just one week after the district court's order of June 28, 1984 in support thereof. After a hearing on both the motion to amend the judgment and the proposed remedies, the district court issued a supplementary order granting the motion to amend the judgment, holding that the plaintiffs could not maintain the action under section 504. On September 27, 1984, the plaintiffs representing the Burke, Coweta and Lee county and the Vidalia City school systems filed their notice of appeal. Record, vol. 40 at 9222-23.

The long history of school desegregation efforts in Georgia is a familiar story and will not be summarized here.3 The allegations of intraschool discrimination in this case are predicated on the widespread practice of grouping students on the basis of ability or achievement and on the assignment of schoolchildren to EMR programs.4 The following descriptions of the local defendants' grouping and EMR program assignment practices are based on the facts found by the district court and, except as set forth infra, are uncontroverted on appeal.5

Neither federal law nor the Georgia State Department of Education mandates or prohibits the grouping of children for instructional purposes in regular classrooms. The decision whether to implement these programs is made by the local school district.

The Vidalia City School District (Vidalia City) first employed achievement grouping in the late 1950's in its eighth grade classrooms, and has since expanded and revised it gradually to include grades one through seven. From the 1970-71 to 1979-80 school terms, Vidalia City used the Scott Foresman reading system to place first graders. Starting in 1980-81 initial placements of first graders in all classes have been based on scores from the Metropolitan Achievement Test (MAT) and recommendations from kindergarten teachers.

Various versions of the Scott Foresman test were employed throughout most of the seventies to group all other elementary school students. In 1977-78, the school district adopted the Holt Basic Reading Series and the Holt Basic Math Series (the "Holt system"). Second and third graders have been organized in reading and math classes according to the Holt system since 1981-82. Students in other classes such as homeroom, science, social studies and health have not been grouped since the 1982-83 school year.

Fourth and fifth grade students have been classified according to the Holt math system since the 1978-79 term. The Holt reading system was implemented during the 1981-82 school year. Children in the sixth and seventh grades have been grouped under the Holt system in math classes since the 1980-81 term and in reading classes since the 1981-82 school year. All other sixth and seventh grade classes are heterogeneous except for social studies, which is organized according to the reading evaluation, and science, which is grouped according to the math evaluation.

Vidalia City does not group students in grades eight through twelve. All of the procedures used include regular reevaluations of a student's ability through end-of-level and end-of-unit tests. Children are not locked or "tracked" into the same group; there is substantial movement of both black and white students upward and downward in achievement levels. See generally Record, vol. 53 at 3162-81, vol. 54 at 3182-3294 (testimony of Dr. Steven Whatley).

The Lee County School District (Lee County) initiated achievement grouping during the 1967-68 school term in the fifth grade and expanded within a few years to all elementary grades. Lee County High School students generally are not classified according to achievement except for those assigned to remedial math and language arts courses in which enrollment is limited to those who have failed or are in danger of failing the high school exit examination. See Record, vol. 55 at 3610 (testimony of Dr. Robert Clay).

Kindergarten students were assigned heterogeneously until the 1982-83 school term when a special developmental kindergarten was added for educationally disadvantaged students. Students are assigned to the developmental kindergarten on the basis of Development Indicators for the Assessment of Learning (DIAL) tests, a preschool readiness examination. See Record, vol. 46 at 1289-90 (testimony of Dr. Robert Calfee) (explaining readiness tests). Within the two kindergartens, students are randomly assigned to classes.

In the Lee County Elementary School, classification decisions are made by a team of school personnel including the principal and the student's teachers. Assignments are based primarily on reading performance as determined by the MacMillan Placement Test, the Georgia Criterion Reference Test (GCRT), the SRA Achievement Test, the MacMillan Skills Test, the Distar Progress Test and the Key Text Progress Test. Grouping decisions are also influenced by oral reading skill, performance in other academic areas, teacher evaluations and practical concerns such as the need to balance class size within each grade.

Class assignments in the Lee County Upper Elementary School are based on SRA composite test results. Teacher recommendations are used in borderline cases. School personnel divide the students into three groups within which the children are assigned randomly.

Lee County regularly monitors each student's performance through end-of-level tests, examinations drafted by the child's own teachers, and standardized tests. Changes in a student's placement may be made at any time during the school year, and decisions on classroom assignments are reassessed at the beginning of each term. Students in Lee County do not remain with the same group of schoolchildren. See generally Record, vol. 55 at 3538-3637 (testimony of Dr. Clay).

The Coweta County School District (Coweta County) was formed in 1969 by the merger of two previously independent segregated school systems and adopted achievement grouping at its inception. Classification of students is limited to grades one through seven and is used for math and language arts classes only. All other classes are heterogeneous.

Placement in the first grade is based on several standardized tests, including the MAT, the Southwestern Regional Lab and Assessment instrument and the GCRT. Student performance and teacher observations are also taken into consideration. In later grades placement is based on performance in the basal reading and math series, see Record, vol. 43 at 568 (testimony of Dr. Phillip McLaughlin), vol. 46 at 1229-30 (testimony of Dr. Calfee) (explaining basal tests), and teacher observations and evaluations.

Coweta County systematically reevaluates students through the Ginn reading and math program, a continuous progress rating system. Additional formal reviews are made every six weeks, at the end of each semester and at the end of the school year. Movement within and between groups is based on teacher observations and evaluations as well as student performance in the classroom and on end-of-level tests. Students are not tracked into any particular group. In high school, achievement grouping is employed only in English classes. See generally Record, vol. 53 at 2934-95 (testimony of Donald Teel).

The Burke County School District (Burke County) classifies achievement by groups in kindergarten through the eighth grade. At the elementary level, grouping encompasses all required academic subjects including English, science, math and social studies. Initial placement is determined by the school principal and counselor and is based on teacher recommendations, standardized test results and basal program performances. The student's instructional reading level receives the greatest weight.

In the junior high school, assignments are made similarly except that students are permitted some flexibility in choosing courses. There is no classification of high school students.

Beginning in the 1982-83 school year, the school system instituted a program of periodic reevaluations of students to determine whether their initial placements remained accurate. Under the current process, student assignments are reassessed after the first six weeks of school. Students do not stay within the same group throughout their elementary school education. See generally Record, vol. 56 at 4177-4203 (testimony of Benny Cannady).

Congress passed the Education of All Handicapped Children Act, 20 U.S.C. Secs. 1400 et seq. (EAHCA) in order to afford all handicapped children6 "a free appropriate public education." See 20 U.S.C. Sec. 1400(c).7 The states have the chief responsibility for accomplishing this goal. See id. Secs. 1412, 1413. Pursuant to this mandate, the State of Georgia has promulgated a variety of regulations establishing and governing its special education program. These regulations provide more specific definitions of the term "handicap" and procedures for placing children in appropriate special education classes. This case concerns the placement procedures for the educably mentally retarded. The plaintiffs do not protest the assignment of students to other educational programs for children with academic or learning difficulties.8

The local defendants implement these regulations through similar procedures. Vidalia City, for instance, operates a two-stage program. The first step involves the review of all proposed special education referrals by a screening committee. Referrals may be made by school personnel, parents or any other person familiar with the child's educational needs. After obtaining parental consent for screening and conducting a number of tests including achievement, intelligence, hearing and vision evaluations, the committee assesses the test results, the child's classroom performance and any remedial efforts already attempted in the regular classroom. If the committee does not perceive a practical alternative, the child is submitted for formal evaluation for special education placement. Parental permission is secured before the formal evaluation, and a variety of tests are conducted including an in-depth psychological evaluation by a certified psychologist or psychometrist. The examinations administered include achievement tests, tests designed to isolate any specific learning disabilities or behavior disorders, and tests measuring adaptive behavior. The adaptive behavior considered is generally the child's in-school conduct, but behavior outside of school is sometimes also evaluated by home visits and other contact with the student's parents.9

After the completion of this process, the placement committee composed of teachers, administrators, psychometrists and others determines whether the child should be assigned to an EMR program. No placements are made without parental consent. All assignments to EMR programs are reexamined once each year.

Lee County's special education program is under the direction of TLW, a tri-county program operated for the benefit of Terrell, Lee and Worth counties which are too small to provide effective services alone. Like Vidalia City, Lee County employs a screening committee to ascertain the child's classroom performance and remedial efforts already attempted in the regular classroom, and to recommend the propriety of referral for a special evaluation. After the referral, the procedures followed are essentially the same as those followed by Vidalia City. The process employed by Coweta County is similar, with a two-step program and reevaluations conducted every three years. Although Burke County uses conferences between the child's parents, teachers and the school principal instead of a more formal screening committee, its procedures are fundamentally the same as those of the other three local defendants. See generally, Record, vol. 48 at 1658-1735 (testimony of Alan White), vol. 49 at 2110-18 (testimony of Dr. Richard Kicklighter), vol. 53 at 3048-3112 (testimony of Marilyn Lamb), vol. 54 at 3297-3334 (testimony of Marlene Tomblin), 3387-3414 (testimony of Barbara Boyer), vol. 55 at 3415-3503 (same).

The plaintiffs first contend that the local defendants' achievement groupings and EMR assignments discriminate against black children on the basis of race in violation of the equal protection clause of the fourteenth amendment. It is undisputed that, in the local defendant school districts, there are more blacks in the lower achievement groups and EMR classes than would be expected from a random distribution, given the percentage of blacks in each school system. Nonetheless, the district court held that (1) the plaintiffs failed to prove that the disparity in the lower achievement groups is the result of the present effects of past discrimination, Record, vol. 39 at 8875, and (2) the local defendants' ability classification schemes serve to remedy the past results of prior segregation through better educational opportunities for black students. Id. at 8884. The district court also held that the plaintiffs failed to show that the racially disparate EMR enrollments are attributable to prior segregation in Georgia schools, that black children are misclassified as EMR more often than white students, or that the defendants had any intent to discriminate against black students. Record, vol. 39 at 8887-89.

On appeal, the plaintiffs challenge only the district court's conclusions respecting the ability grouping, urging that the district court wrongly required them to prove intentional discrimination. They contend that the district court should have presumed the intent to discriminate because of the local defendants' past segregation practices.

It is well-established in this circuit that ability grouping is not per se unconstitutional, even when it results in racial disparity in a school district's classrooms. See Castaneda v. Pickard, 648 F.2d 989, 994 (5th Cir. Unit A 1981); McNeal v. Tate County School District, 508 F.2d 1017, 1020 (5th Cir.1975). In evaluating the constitutional validity of ability grouping, courts have noted the importance of examining the school district's history of segregation.

First, ability grouping, when employed in such traditional circumstances may perpetuate the effects of past discrimination by resegregating, on the basis of ability, students who were previously segregated in inferior schools on the basis of race or national origin. Second, a relatively recent history of discrimination may be probative evidence of a discriminatory motive which, when coupled with evidence of the segregative effect of ability grouping practices, may support a finding of unconstitutional discrimination.

Castaneda, 648 F.2d at 996. With these dangers in mind, we have cautioned that if a school district's ability grouping program has a significant disparate impact on students of different races, the program cannot be implemented until the district operates a unitary school system for a period of at least several years. See, e.g., Castaneda, 648 F.2d at 994; Lemon v. Bossier Parish School Board, 444 F.2d 1400, 1401 (5th Cir.1971) (per curiam); Singleton v. Jackson Municipal Separate School District, 419 F.2d 1211, 1219 (5th Cir.1969) (per curiam) (en banc) (achievement grouping "cannot be employed in any event until unitary school systems have been established"), (case number 28261).10 It is uncontroverted in this case that the local defendants ceased operating dual school systems in 1970-71 and, therefore, are "unitary" in the sense which permits the use of ability grouping under certain circumstances.11

The standard in ascertaining the constitutional propriety of achievement grouping in a unitary school district depends on whether that school system has been declared "fully unitary," or has achieved "unitary status," terms with meanings different from the label "unitary" as used above.12 A school system which was previously segregated can become fully unitary after implementing a constitutionally acceptable desegregation plan and operating in a manner in which "the State does not discriminate between public school children on the basis of race." Lee v. Macon County Board of Education, 584 F.2d 78, 81 (5th Cir.1978); see also Pitts v. Freeman, 755 F.2d 1423, 1426 (11th Cir.1985). To declare a school district as fully unitary and thus terminate a school desegregation case, a district court must first hold a hearing. Id.; see United States v. Texas Education Agency, 647 F.2d 504, 509 (5th Cir. Unit A 1981), cert. denied, 454 U.S. 1143, 102 S. Ct. 1002, 71 L. Ed. 2d 295 (1982). "The plaintiffs should receive notice of the hearing's purpose, and the hearing should give them an opportunity to show why the court should continue to retain jurisdiction." Pitts, 755 F.2d at 1426; see Texas Education Agency, 647 F.2d at 509.

None of the local defendants ever acquired unitary status through this procedure. Although the local defendants were ordered to desegregate in 1970-71 in United States v. Georgia, Civil Action No. 12,972 (N.D.Ga. Dec. 17, 1969) (Lee County, Coweta County and Vidalia City), and Bennett v. Evans, Civil Action No. 1443 (S.D.Ga. June 24, 1970) (Burke County), and have not operated a dual system since, neither United States v. Georgia nor Bennett v. Evans are concluded cases.

Until the local defendants achieve unitary status, they have an affirmative duty to eliminate the consequences of their prior unconstitutional conduct. Pitts, 755 F.2d at 1426. "[O]fficial action that has the effect of perpetuating or reestablishing a dual school system violates [this] duty to desegregate." Id. at 1427 (emphasis in original); see generally Columbus Board of Education v. Penick, 443 U.S. 449, 459-60, 99 S. Ct. 2941, 2947-48, 61 L. Ed. 2d 666, 677-78 (1979); Wright v. Council of the City of Emporia, 407 U.S. 451, 460-62, 92 S. Ct. 2196, 2202-03, 33 L. Ed. 2d 51, 60-61 (1972). Challenges to official action by a local school district which has not achieved unitary status can be successful without proof of discriminatory intent. See Pitts, 755 F.2d at 1427.13 The school system, however, is not obligated to adopt the most desegregative alternative available. See Lee v. Anniston City School System, 737 F.2d 952 (11th Cir.1984).

This circuit's precedent establishes that, despite any resulting numerical racial disproportionality, achievement grouping is permissible in a school district that has not been declared fully unitary "if the school district can demonstrate that its assignment method is not based on the present results of past segregation or will remedy such results through better educational opportunities." McNeal, 508 F.2d at 1020 (emphasis added); see Debra P. by Irene P. v. Turlington, 730 F.2d 1405, 1414 & n. 14 (11th Cir.1984); United States v. Gadsden County School District, 572 F.2d 1049, 1052 (5th Cir.1978) (per curiam); see generally Castaneda, 648 F.2d at 996-97.14 Contrary to the plaintiffs' contention, the district court in this case did not require them to prove discriminatory intent by the local defendants. Instead, the district court purported to apply the McNeal standard, holding that "the plaintiffs have failed to prove that ability grouping, as practiced by the local defendants, and monitored by the state, is the result of the present effects of past intentional discrimination," Record vol. 39 at 8875, and that "ability grouping as practiced by [the] defendants ... is designed to remedy the past results of past segregation through better educational opportunity for the present generation of black students." Id. at 8884. The court found, "the educational tools being utilized by the individual defendants are educationally sound" and "low achieving students are being successfully remediated." Id.

As an alternative to their argument that the district court erroneously required them to prove intent, the plaintiffs claim that the court wrongly determined that the McNeal standard is controlling in this case. According to the plaintiffs, the district court failed to presume the intent to discriminate in conformance with Keyes v. School District No. 1, 413 U.S. 189, 93 S. Ct. 2686, 37 L. Ed. 2d 548 (1973). In Keyes, a case involving the desegregation of the Denver, Colorado schools, the Supreme Court held that once racial disproportionality is proved, the intent to discriminate may be presumed. The widespread race discrimination existing before the time of the suit justified the inference that the disparity was the result of discriminatory purpose which had never been relinquished since the days of segregation by law. This presumption is strong and may be rebutted only through proof by the school district that the intent to discriminate was not a factor contributing to the racial disproportionality. Id., 413 U.S. at 209-11, 93 S. Ct. at 2697-98, 37 L.Ed.2d at 564-65. A racially neutral explanation is insufficient to rebut the presumption. Id., 413 U.S. at 212-14, 93 S. Ct. at 2699-2700, 37 L.Ed.2d at 565-66.

Because the Keyes presumption is founded on a school district's immediate past history of segregation, the plaintiffs do not take the position that it applies directly to the current ability grouping practices of the local defendants. Instead, the plaintiffs note that the local defendants have continuously employed achievement grouping at least since they were first desegregated. Consequently, these earlier assignment plans resulted in racial disproportionality in regular classrooms and must be presumed to be unconstitutional under Keyes. Because the present grouping programs are continuations of the initial schemes, and because ability grouping is allegedly only a "racially neutral explanation," the plaintiffs urge that Keyes requires a finding of intentional discrimination in this case.

Essentially, they claim that ability grouping schemes which result in racial disproportionality cannot be implemented until a school district has operated as a unitary system without such programs for some indefinite period of time. This stance is not without support in this circuit's precedent. In McNeal, we stated that the "rationale of both Singleton and Lemon would bar the use of [achievement grouping] until the district has operated as a unitary system without such assignments for a sufficient period of time to assure that the underachievement of the slower groups is not due to yesterday's educational disparities." McNeal, 508 F.2d at 1020-21 (emphasis added).

The student population in McNeal, however, included black children who had attended inferior segregated schools. The McNeal court's concern was that the prior "lack of educational quality would predictably cause students from the inferior system to immediately be resegregated within the lower classroom sections." Id. at 1020. This reasoning is not pertinent here where the district court found that no student presently in the certified class had ever attended a segregated school. The fact that the grouping of the predecessors to the members of this class may have been influenced by their attendance at segregated schools is not relevant to the constitutional validity of the grouping systems at issue here. As a result, we conclude that the Keyes presumption does not apply and that the district court correctly followed the McNeal standard.

We need not decide whether the district court improperly placed the burden on the plaintiffs to prove that ability grouping resulted from past discrimination, see, e.g., Vaughns v. Board of Education of Prince George's County, 758 F.2d 983, 991 (4th Cir.1985), because we hold that the district court's factual finding, which is unchallenged by the plaintiffs on appeal, that the ability grouping schemes will remedy the consequences of prior segregation through better educational opportunities is not clearly erroneous. See Debra P., 730 F.2d at 1415-16 (applying clearly erroneous standard to similar findings).15 Despite testimony to the contrary, see, e.g., Record, vol. 46 at 1253 (testimony of Dr. Calfee), there is a great deal of evidence in the record indicating that ability grouping is a sound educational practice, see generally, e.g., Record vol. 50 at 2265-2366 (testimony of Dr. Barbara Lerner), 2436-2478 (testimony of Dr. James Kulik), vol. 55 at 3597 (testimony of Dr. Clay), and that disadvantaged children, including black students, in all four local defendant school districts have made significant academic progress since the inception of the achievement grouping programs. See, e.g., Record, vol. 49 at 1991 (testimony of George Griffin), 2071-74 (testimony of William Tidwell), vol. 54 at 3231-37 (testimony of Dr. Whatley), vol. 55 at 3631-35 (testimony of Dr. Clay); Vidalia City Exhibits 22-24; Lee County Exhibits 26-29.16 Therefore, we conclude that the procedures employed by the defendants to assign students did not violate the equal protection clause of the constitution.17

Section 601 of Title VI provides that "[n]o person in the United States shall, on the ground of race, ... be excluded from participation in, be denied the benefits of, or be subjected to discrimination under any program or activity receiving Federal financial assistance." 42 U.S.C. Sec. 2000d (section 601). The plaintiffs assert that this section furnishes an additional basis for relief from the allegedly discriminatory assignment practices of the defendants.

The district court applied disparate impact tests in making its determination under section 601. The defendants argue that the district court should have required the plaintiffs to prove discriminatory intent, citing Castaneda v. Pickard, 648 F.2d 989 (5th Cir. Unit A 1981), where the former Fifth Circuit Court of Appeals held that "discriminatory intent, as well as disparate impact, must be shown in employment discrimination suits brought against public employers under Title VI...." Id. at 1000.

More recent Supreme Court decisions, however, have overruled this aspect of Castaneda. In Guardians Association v. Civil Service Commission, 463 U.S. 582, 103 S. Ct. 3221, 77 L. Ed. 2d 866 (1983), a case involving claims of disproportionate layoffs due to allegedly discriminatory examinations administered to Black and Hispanic police officers, five Justices were of the opinion that the regulations promulgated under Title VI permit the filing of suits alleging a disparate impact theory. See id., 463 U.S. at 607 n. 27, 103 S. Ct. at 3235 n. 27, 77 L. Ed. 2d at 885 n. 27 (summarizing the various positions of the Justices); see also Consolidated Rail Corp. v. Darrone, 465 U.S. 624, ---- n. 9, 104 S. Ct. 1248, 1252 n. 9, 79 L. Ed. 2d 568, 574 n. 9 (1984). Moreover, in Alexander v. Choate, 469 U.S. ----, 105 S. Ct. 712, 83 L. Ed. 2d 661 (1985), a case construing section 504, the Court read Guardians as holding that "actions having an unjustifiable disparate impact on minorities could be redressed through agency regulations ...." Id., 469 U.S. at ----, 105 S. Ct. at 717, 83 L.Ed.2d at 667.18 There is no doubt that the plaintiffs predicated this cause of action on the regulations.19 As a result, the district court correctly applied disparate impact analyses to their Title VI claims.

The elements of a disparate impact claim may be gleaned by reference to cases decided under Title VII, 42 U.S.C. Sec. 2000e et seq. See NAACP v. Medical Center, Inc., 657 F.2d 1322, 1331 (3d Cir.1981) (en banc) (Title VII standards instructive in Title VI case). The plaintiff first must show by a preponderance of the evidence that a facially neutral practice has a racially disproportionate effect, whereupon the burden shifts to the defendant to prove a substantial legitimate justification for its practice. The plaintiff then may ultimately prevail by proferring an equally effective alternative practice which results in less racial disproportionality or proof that the legitimate practices are a pretext for discrimination. See generally Albemarle Paper Co. v. Moody, 422 U.S. 405, 95 S. Ct. 2362, 45 L. Ed. 2d 280 (1975); McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S. Ct. 1817, 36 L. Ed. 2d 668 (1973).

The district court found that the plaintiffs had met their burden of establishing a prima facie case through statistics showing that the racial composition of many of the local defendants' regular classrooms differs from what would be expected from a random distribution. The court also found that the defendants had successfully rebutted the plaintiffs' prima facie case by establishing the educational necessity for grouping students, and that the plaintiffs had failed to show the existence of equally sound educational alternatives to the assignment plans that result in less racial disproportionality. See Record, vol. 39 at 8896-8905. We need not reach the issue whether the district court properly found that the plaintiffs made out an adequate prima facie showing of disparate impact because we agree with the court's other two conclusions.20

The substantial legitimate justification offered to rebut the plaintiffs' prima facie case was the "educational necessity" of the local defendants' grouping practices. This justification is analogous to the "business necessity" exception in Title VII cases. See, e.g., Board of Education of the City School District v. Harris, 444 U.S. 130, 151, 100 S. Ct. 363, 375, 62 L. Ed. 2d 275, 291 (1979) (After determining that a disparate impact analysis could be applied under 20 U.S.C. Sec. 1605(d)(1)(B), the Court stated that, on remand, the burden of rebutting the plaintiffs' prima facie case "perhaps could be carried by proof of 'educational necessity' analogous to ... 'business necessity'...."). Following this guide, we analyze the sufficiency of the defendants' claim of educational necessity in light of cases interpreting the term business necessity.

Griggs v. Duke Power Co., 401 U.S. 424, 91 S. Ct. 849, 28 L. Ed. 2d 158 (1971), disapproved the use of intelligence test scores and a high school diploma requirement as job criteria. The Supreme Court reasoned that "[t]he touchstone is business necessity. If an employment practice which operates to exclude Negroes cannot be shown to be related to job performance, the practice is prohibited." Id., 401 U.S. at 431, 91 S. Ct. at 853, 28 L.Ed.2d at 164. The Court held that the two requirements for employment in that case were not "shown to bear a demonstrable relationship to successful performance of the jobs for which [they were] used." Id. In reaching its conclusion, the Court rejected as insufficient the view that the job criteria "generally would improve the overall quality of the work force," id., and instructed that "[w]hat Congress has commanded is that any tests used must measure the person for the job and not the person in the abstract." Id., 401 U.S. at 436, 91 S. Ct. at 856, 28 L.Ed.2d at 167. The Court has reaffirmed the principle of job relatedness on a number of occasions. See, e.g., Connecticut v. Teal, 457 U.S. 440, 446-47, 102 S. Ct. 2525, 2530, 73 L. Ed. 2d 130, 137 (1982) ("manifest relationship"); Albemarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S. Ct. 2362, 2375, 45 L. Ed. 2d 280, 301 (1975); McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 n. 14, 93 S. Ct. 1817, 1824 n. 14, 36 L. Ed. 2d 668, 678 n. 14 (1973).

Under this precedent, the defendants in this case had the burden of proving that their achievement grouping practices bear a manifest demonstrable relationship to classroom education.21 After reviewing the evidence, the district court concluded that the defendants had "proffered a sufficient nondiscriminatory justification," Record, vol. 39 at 8902, and had "demonstrated that their programs are educationally necessary to accommodate the needs of the student populations in the subject rural Georgia counties." Id. at 8886. The court characterized achievement grouping as an "accepted pedagogical practice," id. at 8899, and found that "the educational need[s] of the students serviced by the instructional groups and compensatory and remedial programs are being addressed by the local defendants' practices." Id. at 8902.

The plaintiffs' disagreement with the district court's holding is twofold. First, they urge that the evidence is insufficient to support a finding that ability grouping per se is educationally desirable. Second, they assign as error the district court's failure to specifically find that the methods used by the local defendants to assign students to achievement groups adequately measure a student's true abilities in each particular class. The plaintiffs point out that any supposed educational advantages gained through ability grouping are negated when, for example, a student is assigned to a group in mathematics as the result of his or her scores on reading tests.

The district court's findings of the educational soundness of interclass ability arrangements per se are not clearly erroneous.22 The record discloses that such grouping permits more resources to be routed to lower achieving students in the form of lower pupil-teacher ratios and additional instructional materials. There is also evidence that ability grouping results in improved class manageability, student and teacher comfort and student motivation. There was expert testimony attesting to the desirability of these practices. See generally, e.g., Record, vol. 50 at 2265-2366 (testimony of Dr. Lerner), 2436-2478 (testimony of Dr. Kulik), vol. 55 at 3597 (testimony of Dr. Clay).23 Furthermore, as noted earlier, the evidence demonstrates that students in the four local defendant school districts achieved higher scores in statewide testing programs since the development of the system and that individual students in the local school districts have improved sufficiently to be reassigned to higher-level achievement groups. See, e.g., Record, vol. 53 at 2958-60 (testimony of Donald Teel), vol. 55 at 3622-24 (testimony of Dr. Clay); Lee County Exhibits 30, 32, 34, 36, 44-46.24

The plaintiffs point to evidence indicating that achievement grouping exacerbates, rather than relieves, the problem of disproportionate lower achievement by black students. See, e.g., Record, vol. 46 at 1270-71 (testimony of Dr. Calfee). This evidence included a series of Analysis of Variance (ANOVA) and multiple regression calculations, and a linear regression analysis. These tests generally suggested that a student's race and the percentage of blacks in a child's class are the most potent predictors of his academic achievement. Citing this evidence, the plaintiffs urged the district court to find that "racial differences in achievement, relied upon by the defendants in making classroom assignments, are in fact caused by the provision of different educational experiences for black and white children." Record, vol. 39 at 8821.

There was other evidence, however, that these statistical studies failed to adequately account for other possible considerations affecting student performance. Dr. Lerner, for example, testified extensively on the factors impacting academic achievement and concluded that "family background" and "hard work," rather than race, had the most powerful and consistent relationships to scholastic success. See generally Record, vol. 50 at 2265-2366 (testimony of Dr. Lerner). We cannot overturn the district court's finding of fact simply because this evidence was merely conflicting. See Western Beef, Inc. v. Compton Investment Co., 611 F.2d 587, 590 (5th Cir.1980).

The plaintiffs next fault the district court's failure to find that the methodology used by the defendants to assign students was valid. We agree that, just as job requirements must adequately measure characteristics related to job performance in the employment context, the criteria by which students are assigned to a specific class must adequately measure the student's abilities in that subject. Cf. Scott v. City of Anniston, 597 F.2d 897, 901-02 (5th Cir.1979) (qualifying tests must be "appropriate for the selection of qualified applicants"), cert. denied, 446 U.S. 917, 100 S. Ct. 1850, 64 L. Ed. 2d 271 (1980).25 Although the district court did not specifically state that the tests employed by the local defendants adequately measure a student's ability in each subject taught in the classes in which the student is grouped, these findings are implicit in the district court's comprehensive opinion. See Rowley v. McMillan, 502 F.2d 1326, 1334 (4th Cir.1974). The court's awareness of the dangers of assignment to one class according to evaluations relevant only to another subject is apparent from its discussion of the grouping models used by the local defendants. See Record, vol. 39 at 8803-05. Because the grounds upon which the court reached its conclusion are evident, the lack of explicit findings does not preclude appellate review. Cf. Grover Hill Grain Co. v. Baughman-Oster, Inc., 728 F.2d 784, 792-93 (6th Cir.1984).

The district court's implied findings are not clearly erroneous. The record reflects that students in all four defendant school districts are grouped in particular classes according to tests, evaluations and performance in the same subject, in closely related subjects or in a broad variety of subjects.26 There is no direct evidence that students are assigned according to criteria not manifestly related to the subject matter taught in the specific class.27 There is evidence in the record establishing the validity of certain of the testing procedures. See, e.g., Record, vol. 47 at 1565-82 (testimony of Dr. Stanley Bernknopf). The reliability of the local defendants' grouping criteria is also supported by the evidence showing improvement in students scores and mobility between achievement groups. Based on these facts and our review of the entire record, the district court did not err in finding that the defendants had rebutted the plaintiffs' prima facie case of disparate impact by establishing the educational necessity of achievement grouping.

The plaintiffs complain that the district court eliminated random assignment accompanied by intraclass grouping as an equally sound educational alternative that results in less racial disproportionality. See Record, vol. 46 at 1293 (testimony of Dr. Calfee) (explaining heterogeneous grouping as a remedy).28 They point to the court's statement that there "is no evidence in the record which supports a conclusion that either homogeneous or heterogeneous grouping is clearly more effective in providing an appropriate education." Record, vol. 39 at 8904. This language, they say, constitutes a recognition by the district court that random grouping would be equally effective as homogeneous achievement grouping.

This stance is based on a misreading of the district court's opinion. The court's statement is only a determination that neither homogeneous nor heterogeneous grouping per se is educationally advantageous. This finding is not clearly erroneous. Although there was evidence to the contrary, see, e.g., Record, vol. 46 at 1297-98, vol. 58 at 4538-39 (testimony of Dr. Calfee), there was expert testimony that integration per se is not educationally advantageous, see Record, vol. 50 at 2288-93, 2301 (testimony of Dr. Lerner), and that homogeneous grouping may be detrimental to higher achieving students. See id. at 2365 (testimony of Dr. Lerner). There was also testimony that intraclass grouping is not as beneficial as interclass grouping. See, e.g., id. at 2303-04 (testimony of Dr. Lerner). As noted previously, the district court determined that interclass achievement grouping is educationally beneficial. Based on these considerations, the district court was not clearly erroneous in finding that random assignment and intraclass grouping is not an equally sound educational alternative.29

The nature of the plaintiffs' special education claims are far from clear. They acknowledge that the state regulations, if properly implemented and interpreted by the local defendants, would not violate Title VI. Brief of Appellants at 54.30 Rather, they appear to assert that the defendants' 1) violation of the procedural regulations, and 2) misinterpretation of the state I.Q. score regulation contravene section 601.31

Correctly applying a disparate impact analysis, the district court held that the plaintiffs successfully proved a prima facie case. However, the court found that the defendants successfully rebutted this showing and that the plaintiffs failed to establish the existence of any equally effective alternative practices which would result in less racial malapportionment.

Generally, to establish a prima facie case of disparate impact based on race the plaintiffs must show that the defendants' racially neutral practice detrimentally affects persons of a particular race to a greater extent than other races. See generally Rowe v. General Motors Corp., 457 F.2d 348, 354 (5th Cir.1972); B. Schlei & P. Grossman, Employment Discrimination Law 162 (2d Ed.1983). "In making a prima facie case in a disparate impact suit, ... the plaintiff must not merely prove circumstances raising an inference of discriminatory impact; he must prove the discriminatory impact at issue." Johnson v. Uncle Ben's, Inc., 657 F.2d 750, 753 (5th Cir.1981), cert. denied, 459 U.S. 967, 103 S. Ct. 293, 74 L. Ed. 2d 277 (1982). Consequently, the plaintiffs must show that the practices of the local defendants impacted more harshly on black children than on other students.

The plaintiffs furnished statistical evidence establishing with a high level of mathematical confidence that the percentage of black students assigned to EMR programs in each defendant school district was significantly greater than what would be expected from a random distribution of schoolchildren, given the percentage of black children in the district's student population as a whole. These statistics were derived from information about the number of black and white students in the defendant school district's EMR classes and student populations. See, e.g., Record, vol. 41 at 41-49 (testimony of Dr. Martin Shapiro); Plaintiffs' Exhibits 778, 883, 895-97.

This evidence is insufficient to establish a prima facie case of disparate impact with respect to either of the practices of the local defendants. To assess the effect of the local defendants' first practice, the total pool of students against which black children should have been measured is not the entire student body of the school district but the group of students who have been classified as EMR through a process not complying with the applicable regulations. However, the plaintiffs offered no evidence that the percentage of black students in this pool was greater than the percentage which would be expected from a random distribution. The number of white children assigned to EMR programs through the same impermissible application of the procedural regulations was not examined by the plaintiffs. In fact, the only evidence introduced at the trial indicated that procedural violations occur when evaluating white students at a rate similar to that when assigning black schoolchildren. See Record, vol. 43 at 684-86 (testimony of Dr. McLaughlin).32

The evidence is similarly insufficient to establish a prima facie case of discrimination because of the flexible application of I.Q. scores. The plaintiffs argue that the local defendants should have interpreted the state I.Q. score regulation as prohibiting the classification of students as EMR when their I.Q.'s were higher than two standard deviations below the mean of the test employed. See generally, infra Part V(A)(2). Although the plaintiffs allege that the practice of misinterpreting the I.Q. score regulation has a disparate impact on black students, they offered no evidence that the percentage of black schoolchildren in the group of all children with I.Q.'s greater than two standard deviations below the mean of the test employed who had been assigned to EMR programs was greater than the percentage which would be expected from a random distribution. The plaintiffs did not consider the number of white children who were classified as EMR with similar I.Q. scores.

The plaintiffs simply failed to meet their burden of showing that the criticized practices here did not affect white students to the same degree as black students. Practices which detrimentally affect all groups equally do not have a discriminatory effect. The plaintiffs' dogged reliance on specific instances of alleged misclassifications of black students is meaningless to the Title VI issue in the absence of evidence of the frequency of misclassifications of white schoolchildren. The plaintiffs' statistics establish that black children are assigned to EMR classes at a disproportionate rate but do not demonstrate that the disparity is attributable to the local defendants' practice of violating or misinterpreting regulations.33 We hold that the evidence presented did not, as a matter of law, establish a prima facie case of disparate impact.34

Section 504 of the Rehabilitation Act of 1973, 29 U.S.C. Sec. 794, proscribes discrimination against handicapped persons in federal programs or activities receiving federal financial assistance:

No otherwise qualified handicapped individual ... shall, solely by reason of his handicap, be excluded from the participation in, be denied the benefits of, or be subjected to discrimination under any program or activity receiving Federal financial assistance or under any program or activity conducted by an Executive agency or by the United States Postal Service.

29 U.S.C. Sec. 794. To make out a case under this section, the plaintiffs charge (1) that the local defendants contravened a number of federal regulations in operating their EMR assignment programs and (2) that black children are misclassified as educably mentally retarded in a manner constituting handicap discrimination.

Federal regulations passed pursuant to the Rehabilitation Act provide a number of procedural safeguards to ensure that students are properly classified as EMR program candidates. The regulations require a full individual evaluation of the child before placement in any EMR program. See 34 C.F.R. Sec. 104.35(a). The evaluation team must be composed of persons knowledgeable about the child, the meaning of the evaluation data and the placement options. See 34 C.F.R. Sec. 104.35(c)(3). In addition, the school must ensure the appropriate parental consent to and participation in the evaluation process. See 34 C.F.R. Sec. 104.36. The appraisal of the child's eligibility must be based on validated tests and other material tailored to assess specific areas of educational need, see 34 C.F.R. Sec. 104.35(b)(1), (2), and must draw on a variety of information including the results of aptitude and achievement tests, teacher recommendations, and the child's physical condition, social and cultural background, and adaptive behavior. See 34 C.F.R. Sec. 104.35(c)(1).

The team must also determine the least restrictive environment available that will meet the child's educational needs. See C.F.R. Secs. 104.34, 104.35(c)(4). All this information must be considered and documented, see 34 C.F.R. Sec. 104.35(c)(2), and placement decisions are required to be reevaluated periodically. See 34 C.F.R. Sec. 104.35(d).

Initially, the district court found that all four local school districts had violated a number of these regulations. See Record, vol. 39 at 8989-90, 8993, 8998-99, 9015-17. The violations related to such matters as the failure to administer psychological exams, assess adaptive behavior, conduct timely placement reevaluations, consider less restrictive alternatives to EMR placement and ensure adequate parental participation in the placement process. The local defendants do not deny these violations, claiming only that they are attributable to human error and financial and administrative difficulties.

In its supplementary order the district court held that the Supreme Court's decision in Smith v. Robinson, 468 U.S. ----, 104 S. Ct. 3457, 82 L. Ed. 2d 746 (1984), precludes any relief under section 504 for these regulatory violations. According to the district court, since the procedural violations were actionable to the same extent under the EAHCA, the plaintiffs were required under Robinson to proceed under the EAHCA and first exhaust their state administrative remedies. The court reasoned that the plaintiffs admittedly sought relief only under section 504 and did not pursue their EAHCA administrative avenues to their finality and consequently were entitled to no section 504 relief.

In Robinson, the plaintiff alleged that a child had been denied special education placement in violation of state law, section 504, the EAHCA and the fourteenth amendment. The plaintiff prevailed under state law and applied for attorney's fees pursuant to 42 U.S.C. Sec. 1988 and 29 U.S.C. Sec. 794a(b). The Supreme Court held that when the EAHCA overlaps with more general civil rights provisions such as section 504 and the fourteenth amendment, the more specific EAHCA "is the exclusive avenue through which the child and his parents or guardian can pursue their claim." Robinson, 468 U.S. at ----, 104 S. Ct. at 3470, 82 L.Ed.2d at 766. In addressing the relationship between the EAHCA and section 504, the court stated

there is no doubt that the remedies, rights, and procedures Congress set out in the [EAHCA] are the ones it intended to apply to a handicapped child's claim to a free appropriate public education. We are satisfied that Congress did not intend a handicapped child to be able to circumvent the requirements or supplement the remedies of the [EAHCA] by resort to the general antidiscrimination provision of [section] 504.

There is no suggestion that [section] 504 adds anything to petitioners' substantive right to a free appropriate public education. The only elements added by [section] 504 are the possibility of circumventing the [EAHCA] administrative procedures and going straight to court with a [section] 504 claim, the possibility of a damages award in cases where no such award is available under the [EAHCA], and attorney's fees.

Id. 468 U.S. at ---- - ----, 104 S. Ct. at 3473-74, 82 L.Ed.2d at 669-70 (footnotes omitted).

The plaintiffs make three arguments why Robinson does not bar relief for the regulatory violations. First, they point out that the Supreme Court established an exception to the exclusivity rule when "the [EAHCA] is not available or where [section] 504 guarantees substantive rights greater than those available under the [EAHCA]." Id., 468 U.S. at ----, 104 S. Ct. at 3474, 82 L.Ed.2d at 771. They contend that their complaint that normal children were misdiagnosed as being retarded is not actionable under the EAHCA, which merely guarantees the rights of persons properly characterized as handicapped.

While it may be true that the EAHCA does not encompass the plaintiffs' substantive claim of handicap discrimination, all the violations of the Rehabilitation Act regulations are also violations of federal and state regulations promulgated pursuant to the EAHCA. See, e.g., 34 C.F.R. Sec. 300.531 and State Exhibit 158, Georgia Department of Education Regulations and Procedures, Program for Exceptional Children IDDFd3-8, Sec. III(B)(4) (requiring full individual evaluation before placement) [hereinafter cited as "Georgia Regulation"]; 34 C.F.R. Secs. 300.532(e), 300.533(a)(3), 300.540 and Georgia Regulation, IDDFd3-8 to IDDFd3-9, Secs. III(B)(1), III(C)(1)(c) (composition of evaluation team); 34 C.F.R. Secs. 300.345, 300.502--300.505, 300.508(b), 300.561, 300.562(b)(3), 300.564, 300.567(a), 300.569(b), (c), 300.571 and Georgia Regulation, IDDFd3 to IDDFd3-6, Secs. I(A)--(E) (parental rights and participation); 34 C.F.R. Sec. 533(a)(1) and Georgia Regulation, IDDFd3-9, Sec. III(C)(1)(a) (information to be considered); 34 C.F.R. Secs. 300.550--330.556 and Georgia Regulation IDDFd3-7 to IDDFd3-9, Secs. II(A)--(C), III(C)(1)(d) (least restrictive environment); 34 C.F.R. Sec. 533(2) and Georgia Regulation, IDDFd3-9, Sec. III(C)(1)(b) (requirement that information be documented); 34 C.F.R. Sec. 534(b) and Georgia Regulation, IDDFd3-9, Sec. III(B)(12) (periodic reevaluations); see generally 34 C.F.R. Sec. 104.36 (compliance with EAHCA procedures ensures compliance with Rehabilitation Act regulations). The plaintiffs, therefore, cannot avoid the exclusivity rule for this reason.

Second, the plaintiffs contend that Robinson is inapplicable because their claim falls within a second exception to the exclusivity principle which applies to challenges to state "procedures." However, the Robinson Court noted only that "maintenance of an independent due process challenge to state procedures would not be inconsistent with the [EAHCA]'s comprehensive scheme." Robinson, 468 U.S. at ---- n. 17, 104 S. Ct. at 3470 n. 17, 82 L. Ed. 2d at 766 n. 17. Because the plaintiffs asserted no constitutional challenge to the validity of the state procedures, this exception is of no aid to them. Cf. Manecke v. School Board of Pinellas County, 762 F.2d 912 (11th Cir.1985) (due process challenge upheld); Rose v. State of Nebraska, 748 F.2d 1258, 1263-64 (8th Cir.1984) (same).

Finally, the plaintiffs urge that it would have been futile for them to pursue their EAHCA administrative remedies. The Robinson Court appeared to recognize that the exhaustion of EAHCA administrative remedies may not be required if doing so would be an exercise in futility. See Robinson, 468 U.S. at ---- n. 22, 104 S. Ct. at 3473 n. 22, 82 L. Ed. 2d at 770 n. 22. The district court observed that at the time the plaintiffs filed suit, the administrative procedures used by the State of Georgia pursuant to the requirements of the EAHCA were unlawful, see Helms v. McDaniel, 657 F.2d 800 (5th Cir. Unit B 1981), cert. denied, 455 U.S. 946, 102 S. Ct. 1443, 71 L. Ed. 2d 658 (1982), and, therefore, that the plaintiffs might not have been required to pursue them.35 However, the court reasoned that the suspension of the exhaustion requirement did not affect the exclusivity of the EAHCA and that the plaintiffs were still required by Robinson to sue under the EAHCA rather than section 504.36

We disagree with the district court's conclusion that the EAHCA provides the exclusive means for raising the procedural violations found in this case. The Robinson Court held only that a plaintiff who had prevailed under state law and alleged substantial claims under section 504 and the EAHCA could not recover attorney's fees under the Rehabilitation Act. The Court, therefore, was not confronted with whether a plaintiff who has sued only under section 504 is precluded from all relief when the EAHCA provides equivalent substantive rights.

The Robinson Court based its holding on Congress' desire to prevent the circumvention of the EAHCA's comprehensive administrative scheme and on the legislature's intent that a plaintiff not be allowed to "supplement the remedies of the [EAHCA] by resort to ... [section] 504." Robinson, 468 U.S. at ----, 104 S. Ct. at 3472, 82 L.Ed.2d at 769. Because, as found by the district court, administrative exhaustion would have been futile in this case, the first concern is not relevant here. Although the goal of limiting the remedies available to the plaintiff would be thwarted if all section 504 relief were allowed in this case, we see no reason why the district court, on remand, could not limit the benefits of section 504 to that available under both that statute and the EAHCA, thereby satisfying the statutory purpose to restrict access to the EAHCA by resort to section 504. A contrary view would require the plaintiffs to engage in the procedural gymnastics of amending their complaint by relabeling the lawsuit as an EAHCA cause of action. Congress could have hardly intended such a result, particularly where, as here, the violations of the regulations are largely conceded by the defendants. We hold that under the circumstances of this case, Robinson does not preclude section 504 relief which does not exceed that available under the EAHCA.

Federal regulations require that "significantly subaverage general intellectual functioning" and "deficits in adaptive behavior," 34 C.F.R. Sec. 300.5(b)(4), be identified in a child before placing the student in an EMR program. In accordance with the federal mandate, the Georgia regulation interprets "mildly mentally handicapped," the first level of the classification "mentally handicapped" which identifies persons eligible for EMR programs, in terms of intellectual functioning and adaptive behavior.

Intellectual functioning--performance on an individually administered psychological examination shall range between two and three standard deviations below the mean of the test utilized.

Adaptive behavior--significant deficits must be present as defined and measured by at least one standardized instrument. This information should be supported by clinical judgment, teacher and parent observations, case study information and/or informal evaluation instrument(s).

Georgia Regulation, IDDFd3-18 to 19, Secs. IV(A)(2)(a)(1), (2).

The plaintiffs insist that the local defendants misinterpreted and misapplied these two provisions. In their view, the "intellectual functioning" criterion establishes a strict rule that students with I.Q.'s above two standard deviations below the mean of the test used cannot be assigned to EMR programs. They also urge that the "adaptive behavior" provision requires that out-of-school adaptive behavior be considered and weighed the same as in-school adaptive behavior. Because the school districts permit flexibility in applying the I.Q. criterion and frequently do not consider out-of-school adaptive behavior, the plaintiffs claim that they violate the state regulations and, consequently, the federal regulations they interpret, resulting in the misclassification of a number of black children.

The district court originally held that the state I.Q. criterion permitted flexibility, but only within a specific limited range measured by the standard error of measurement as explained by the American Association on Mental Deficiency (AAMD) in its Manual on Classification in Mental Retardation. See State Exhibit 292, Grossman, Classification in Mental Retardation 22-25 (1983). The court concluded that "the practices as defined and endorsed by the AAMD evidence best professional practices [and] ... must be incorporated into the state's and local school districts' special education evaluation and placement policies." Record, vol. 39 at 8930. The court declined to order that the local defendants exceed the requirements of the existing state and federal regulations and consider out-of-school adaptive behavior, crediting testimony that "the school setting is the more important evaluation tool inasmuch as the ultimate [determination] is an educational one." Id. at 8931.

The district court's interpretations of the state regulations are not erroneous. The court's construction of the I.Q. score regulation was based on the factual findings that including a standard error of measurement is sound and that the range suggested by the AAMD is professionally desirable. Although the state regulation does not explicitly refer to the standard error of measurement, a number of experts testified at the trial that inclusion of this amount of flexibility in considering I.Q. scores is necessary for a meaningful numerical cutoff. See, e.g., Record, vol. 49 at 2126-28, 2135 (testimony of Dr. Kicklighter) (standard error of measurement is an intrinsic part of I.Q. test). Furthermore, there is substantial evidence in the record supporting the view that the AAMD guidelines are acceptable professional tools. See, e.g., State Exhibit 292; Record vol., 43 at 443-46, 644-45, 670-81 (testimony of Dr. McLaughlin).37

The district court's interpretation of the state adaptive behavior regulation was premised on the factual findings that evaluating a student's out-of-school adaptive behavior is not always possible "[f]rom a practical and financial standpoint," and that in-school adaptive behavior "is the more important evaluative tool." Record, vol. 39 at 8931. These findings are not clearly erroneous. Although there was evidence to the contrary, see, e.g., Record, vol. 43 at 665-66 (testimony of Dr. McLaughlin), vol. 44 at 885-86 (testimony of Dr. Katherine Blake), together with information that school officials on occasion did visit the children's homes to assess adaptive behavior, testimony from two expert witnesses for the defendants, Dr. Kicklighter and Dr. Reschley, confirmed that in-school adaptive behavior is the more important criterion. See, e.g., Record, vol. 49 at 2145-48 (testimony of Dr. Kicklighter), vol. 52 at 2781 (testimony of Dr. Reschley).38 In fact, the district court specifically credited Dr. Reschley's testimony on this point. See id., vol. 39 at 8910, 8931. As found by the district court, there was also substantial evidence that evaluating out-of-school adaptive behavior is often impractical. In light of these factual findings, supported by the record, the district court's conclusion was correct.

The district court treated the issue concerning the interpretation of the regulations as relevant both to the plaintiffs' assertion that the defendants had violated the regulations promulgated under section 504, and to the substantive handicap discrimination count. See Record, vol. 40 at 921 n. 3. In its supplementary order, the court held that to the extent the misinterpretation of the I.Q. score regulation was relevant to the regulatory violations, it was barred by Robinson. However, because the section 504 regulatory violation also constitutes an infringement of an EAHCA regulation, see 34 C.F.R. Sec. 104.3(k)(2)(iii) (defining "qualified handicapped person" in terms of EAHCA definition), under the reasoning of the previous section, we hold that Robinson does not preclude all relief under section 504 for the misinterpretation of the I.Q. score regulation.

The plaintiffs' substantive handicap discrimination claim is predicated on the existence of widespread misclassifications of normal black students as mentally retarded, resulting in an oversubscription of black children in state EMR programs. The district court correctly recognized that section 504 encompasses handicap discrimination resulting from misclassifications.39 See 34 C.F.R. Sec. 104, Appendix A, p. 374 (regulatory procedures are "designed to ensure that children are not misclassified, unnecessarily labeled as being handicapped, or incorrectly placed because of inappropriate selection, administration, or interpretation of evaluation materials"); Carter v. Orleans Parish Public Schools, 725 F.2d 261, 262-63 (5th Cir.1984) (per curiam); Larry P. v. Riles, 495 F. Supp. 926, 967-68 (N.D.Cal.1979). The court, however, was "unable to conclude that certain students or classes of students excluded from regular classrooms were in fact ... not mentally retarded," Record, vol. 39 at 8926 n. 41, and found that the "plaintiffs failed to demonstrate any direct evidence of intentional misclassification." Id. at 8925. These determinations of fact are reviewable under the clearly erroneous rule.

We begin our analysis by focusing on the scope of section 504. This circuit has acknowledged that section 504 affords relief for disparate impact as well as intentional discrimination. See Stutts v. Freeman, 694 F.2d 666, 669 (11th Cir.1983); Prewitt v. United States Postal Service, 662 F.2d 292 (5th Cir. Unit B 1981). Because the plaintiffs have requested compensatory damages in the form of remediation, see Powell v. DeFore, 699 F.2d 1078, 1081 (11th Cir.1983) (per curiam), the defendants urge that intentional discrimination must be shown. Some courts have held that a plaintiff must prove bad faith or the intent to discriminate on the basis of handicap in order to recover damages under section 504. E.g., Carter, 725 F.2d at 264; Scokin v. State of Texas, 723 F.2d 432, 441 (5th Cir.1984). This circuit recently deferred the question, describing the issue of the remedies available under section 504 as "murky." Manecke, 762 F.2d at 921 n. 8.

We conclude that, in this case, the plaintiffs must prove intentional discrimination or bad faith in order to be entitled to relief under section 504. We need not decide whether a showing of disparate impact is sufficient to permit the recovery of damages because we hold that the disparate impact cause of action authorized by the Rehabilitation Act regulations, see 34 C.F.R. Sec. 104.4(b)(4), cannot meaningfully be applied to the misclassification allegations in this case. A prerequisite to employing a disparate impact analysis is the existence of readily discernible classifications of people which permit an evaluation of possible different situations and treatment of the groups. Here the proper identification of the classes themselves is at issue, thus making it impossible to adequately probe the differing treatment of various groups. The plaintiffs have cited no cases, and we are aware of none, in which a disparate impact theory was employed to prove handicap discrimination based on the erroneous classification of normal people. See generally Garrity v. Gallen, 522 F. Supp. 171, 206 (D.N.H.1981) (discussing difficulties in adapting typical Title VI discrimination model to the problem of handicap discrimination). As the regulations state, a section 504 disparate impact claim must be based on criteria or methods that "have the effect of subjecting qualified handicapped persons to discrimination on the basis of handicap." 34 C.F.R. Sec. 104.4(b)(4) (emphasis added).

The district court's finding that the local defendants did not intentionally misclassify any students as being "mentally retarded" is not clearly erroneous.40 A student is "misclassified" for the purpose of this review if he is assigned to an EMR program when he does not meet the definition of "mentally retarded." The plaintiffs claim that a child should be presumed to be misclassified if he is assigned to an EMR program through a process not complying with the state and federal standards. We know of no authority to support this proposition.41 Although noncompliance with procedural safeguards may increase the risk of misclassification, it cannot be said to conclusively establish it. We will not assume, for example, that a child assigned to an EMR program is actually normal simply because his parents did not participate in the evaluation as required by federal regulations. The general rule has always been that the plaintiffs bear the burden of proof in a section 504 handicap discrimination case. See, e.g., NAACP v. Medical Center, Inc., 657 F.2d 1322, 1333 (3d Cir.1981) (en banc) (as amended) ("The ultimate burden of persuasion on the issue of illegal discrimination always remains with the plaintiffs.").

To support their allegations of intentional misclassification, the plaintiffs point to a variety of evidence including the local defendants' admitted regulatory violations, the undisputed oversubscription of black students in EMR programs, studies suggesting the existence of misclassifications and alleged individual examples of misassignments. However, the record shows that the local defendants' interpretations of the applicable regulations were reasonable,42 ] and that the admitted regulatory violations were the result of human error and administrative and financial difficulties. Although certain of the plaintiffs' statistics suggest the existence of misclassifications, there is no direct evidence that the arguable misclassifications were intentional or done in bad faith.43 The studies concentrated on the likelihood of misclassification under the present methods of evaluation rather than on the possibility of intentional discrimination. See, e.g., State Exhibit 294, K. Heller, W. Holtzman & S. Messick, Placing Children in Special Education: A Strategy for Equity 3-74, 182-230 (1982). All of the examples of allegedly misclassified children are, as noted by the district court, "borderline" cases.44 Although the record reveals that evaluation teams assigned certain children to EMR programs without considering all the evidence required by the regulations and updating that information in accordance with the law, the plaintiffs did not submit additional data to the district court establishing that all these candidates were not actually mentally retarded.45 Based on these facts and the other evidence, we cannot say that the district court's finding that any misclassifications were not intentional or made in bad faith is clearly erroneous. We affirm the district court's dismissal of the plaintiffs' substantive handicap discrimination claim.

We affirm the judgment of the district court in its entirety except the dismissal of the plaintiffs' claim that the defendants violated a number of procedural regulations passed pursuant to the Rehabilitation Act and misinterpreted the state I.Q. score regulation. On remand, the district court is directed to grant appropriate relief on these claims which does not exceed that available under the EAHCA.46

AFFIRMED in part, REVERSED in part and REMANDED with directions.