with whom Mr. Justice Marshall joins, dissenting.
The Court holds that the job qualification examination (Test 21) given by the District of Columbia Metropolitan Police Department does not unlawfully discriminate on the basis of race under either constitutional or statutory standards.
*257Initially, it seems to me that the Court should not pass on the statutory questions, because they are not presented by this case. The Court says that respondents' summary judgment motion “rested on purely constitutional grounds,” ante, at 236, and that “the Court of Appeals erroneously applied the legal standards applicable to Title VII cases in resolving the constitutional issue before it,” ante, at 238. There is a suggestion, however, that petitioners are entitled to prevail because they met the burden of proof imposed by 5 U. S. C. § 3304. Ante, at 249 n. 15. As I understand the opinion, the Court therefore holds that Test 21 is job related under § 3304, but not necessarily under Title VII. But that provision, by the Court’s own analysis, is no more in the case than Title VII; respondents’ “complaint asserted no claim under § 3304.” Ante, at 234 n. 2. Cf. ante, at 238 n. 10. If it was “plain error” for the Court of Appeals to apply a statutory standard to this case, as the Court asserts, ante, at 238-239, then it is unfortunate that the Court does not recognize that it is also plain error to address the statutory issues in Part III of its opinion.
Nevertheless, although it appears unnecessary to reach the statutory questions, I will accept the Court’s conclusion that respondents were entitled to summary judgment if they were correct in their statutory arguments, and I would affirm the Court of Appeals because petitioners have failed to prove that Test 21 satisfies the applicable statutory standards.1 All parties’ arguments and *258both lower court decisions were based on Title VII standards. In this context, I think it wrong to focus on § 3304 to the exclusion of the Title VII standards, particularly because the Civil Service Commission views the job-relatedness standards of Title VII and § 3304 as identical.2 See also infra, at 263.
In applying a Title VII test,3 both the District Court and the Court of Appeals held that respondents had offered sufficient evidence of discriminatory impact to shift to petitioners the burden of proving job relatedness. 348 F. Supp. 15, 16; 168 U. S. App. D. C. 42, 45-47, 512 F. 2d 956, 959-961. The Court does not question these rulings, and the only issue before us is what petitioners were required to show and whether they carried their burden. The Court agrees with the District Court’s conclusion that Test 21 was validated by a positive relationship between Test 21 scores and performance in police training courses. This result is based upon the Court’s reading of the record, its interpretation of in*259structions governing testing practices issued by the Civil Service Commission (CSC), and “the current views of the Civil Service Commissioners who were parties to the case.” We are also assured that today’s result is not foreclosed by Griggs v. Duke Power Co., 401 U. S. 424 (1971), and Albemarle Paper Co. v. Moody, 422 U. S. 405 (1975). Finally, the Court asserts that its conclusion is “the much more sensible construction of the job-relatedness requirement.” Ante, at 251.
But the CSC instructions cited by the Court do not support the District Court’s conclusion. More importantly, the brief filed in this Court by the CSC takes the position that petitioners did not satisfy the burden of proof imposed by the CSC guidelines. It also appears that longstanding regulations of the Equal Employment Opportunity Commission (EEOC) — previously endorsed by this Court — require a result contrary to that reached by the Court. Furthermore, the Court’s conclusion is inconsistent with my understanding of the interpretation of Title VII in Griggs and Albemarle. I do not find this conclusion “much more sensible,” and with all respect I suggest that today’s decision has the potential of significantly weakening statutory safeguards against discrimination in employment.
I
On October 12, 1972, the CSC issued a supplement to the Federal Personnel Manual containing instructions for compliance with its general regulations concerning employment practices.4 The provision cited by the Court *260requires that Test 21 “have a demonstrable and rational relationship to important job-related performance objectives identified by management.” “Success in training” is one example of a possible objective. The statistical correlation established by the Futransky validity study, ante, at 251 n. 17, was between applicants’ scores on Test 21 and recruits’ average scores on final examinations given during the police training course.
It is hornbook law that the Court accord deference to the construction of an administrative regulation when that construction is made by the administrative authority responsible for the regulation. E. g., Udall v. Tollman, 380 U. S. 1, 16 (1965). It is worthy of note, therefore, that the brief filed by the CSC in this case interprets the instructions in a manner directly contrary to the Court, despite the Court’s claim that its result is supported by the Commissioners’ “current views.”
“Under Civil Service Commission regulations and current professional standards governing criterion-related test validation procedures, the job-relatedness of an entrance examination may be demonstrated by proof that scores on the examination predict properly measured success in job-relevant training (regardless of whether they predict success on the job itself).
“The documentary evidence submitted in the district court demonstrates that scores on Test 21 are predictive of Recruit School Final Averages. There *261is little evidence, however, concerning the relationship between the Recruit School tests and the substance of the training program, and between the substance of the training program and the post-training job of a police officer. It cannot be determined, therefore, whether the Recruit School Final Averages are a proper measure of success in training and whether the training program is job-relevant.” Brief for CSC 14-15 (emphasis added).
The CSC maintains that a positive correlation between scores on entrance examinations and the criterion of success in training may establish the job relatedness of an entrance test — thus relieving an employer from the burden of providing a relationship to job performance after training — but only subject to certain limitations.
“Proof that scores on an entrance examination predict scores on training school achievement tests, however, does not, by itself, satisfy the burden of demonstrating the job-relatedness of the entrance examination. There must also be evidence — the nature of which will depend on the particular circumstances of the case — showing that the achievement test scores are an appropriate measure of the trainee's mastery of the material taught in the training program and that the training program imparts to a new employee knowledge, skills, or abilities required for performance of the post-training job.” Id., at 24-25.
Applying its standards5 the CSC concludes that none of *262the evidence presented in the District Court established “the appropriateness of using Recruit School Final Averages as the measure of training performance or the relationship of the Recruit School program to the job of a police officer.” Id., at 30.6
The CSC’s standards thus recognize that Test 21 can be validated by a correlation between Test 21 scores and recruits’ averages on training examinations only if (1) the training averages predict job performance or (2) the averages are proved to measure performance in job-related training. There is no proof that the recruits’ average is correlated with job performance after completion of training. See n. 10, infra. And although a positive relationship to the recruits’ average might be sufficient to validate Test 21 if the average were proved to reflect mastery of material on the training curriculum that was in turn demonstrated to be relevant to job performance, the record is devoid of proof in this regard. First, there is no demonstration by petitioners that the training-course examinations measure comprehension of the training curriculum; indeed, these examinations do not even appear in the record. Furthermore, the Futran-sky study simply designated an average of 85 on the *263examination as a “good” performance and assumed that a recruit with such an average learned the material taught in the training course.7 Without any further proof of the significance of a score of 85, and there is none in the record, I cannot agree that Test 21 is predictive of “success in training.”
II
Today’s decision is also at odds with EEOC regulations issued pursuant to explicit authorization in Title YII. 42 U. S. C. § 2000e-12 (a). Although the dispute in this case is not within the EEOC’s jurisdiction, as I noted above, the proper construction of Title VII nevertheless is relevant. Moreover, the 1972 extension of Title VII to public employees gave the same substantive protection to those employees as had previously been accorded in the private sector, Morton v. Mancari, 417 U. S. 535, 546-547 (1974), and it is therefore improper to maintain different standards in the public and private sectors. Chandler v. Roudebush, 425 U. S. 840, 864 (1976). See n. 2, supra.
As with an agency’s regulations, the construction of a statute by the agency charged with its administration is entitled to great deference. Traficante v. Metropolitan Lije Ins. Co., 409 U. S. 205, 210 (1972); Udall v. Tallman, 380 U. S., at 16; Power Reactor Co. v. Electricians, 367 U. S. 396, 408 (1961). The defer*264ence. due the pertinent EEOC regulations is enhanced by the fact that they were neither altered nor disapproved when Congress extensively amended Title VII in 1972.8 Chemehuevi Tribe of Indians v. FPC, 420 U. S. 395, 410 (1975); Cammarano v. United States, 358 U. S. 498, 510 (1959); Allen v. Grand Central Aircraft Co., 347 U. S. 535, 547 (1954); Massachusetts Mut. Life Ins. Co. v. United States, 288 U. S. 269, 273 (1933). These principles were followed in Albe-marle — where the Court explicitly endorsed various regulations no fewer than eight times in its opinion, 422 U. S., at 431-4369—and Griggs, 401 U. S., at 433-434.
The EEOC regulations require that the validity of a job qualification test be proved by “empirical data demonstrating that the test is predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are being evaluated.” 29 CFR § 1607.4 (c) (1975). This construction of Title YII was approved in Albemarle, where we quoted this provision and remarked that “[t]he message of these Guidelines is the same as that of the Griggs case.” 422 TJ. S., at 431. The regulations also set forth minimum standards for *265validation and delineate the criteria that may be used for this purpose.
“The work behaviors or other criteria of employee adequacy which the test is intended to predict or identify must be fully described; and, additionally, in the case of rating techniques, the appraisal form(s) and instructions to the rater(s) must be included as a part of the validation evidence. Such criteria may include measures other than actual work proficiency, such as training time, supervisory ratings, regularity of attendance and tenure. Whatever criteria are used they must represent major or critical work b&haviors as revealed by careful job analyses.” 29 CFR § 1607.5 (b) (3) (1975).
This provision was also approved in Albemarle. 422 U. S., at 432, and n. 30.
If we measure the validity of Test 21 by this standard, which I submit we are bound to do, petitioners’ proof is deficient in a number of ways similar to those noted above. First, the criterion of final training examination averages does not appear to be “fully described.” Although the record contains some general discussion of the training curriculum, the examinations are not in the record, and there is no other evidence completely elucidating the subject matter tested by the training examinations. Without this required description we cannot determine whether the correlation with training examination averages is sufficiently related to petitioners’ need to ascertain “job-specific ability.” See Albemarle, 422 U. S., at 433. Second, the EEOC regulations do not expressly permit validation by correlation to training performance, unlike the CSC instructions. Among the specified criteria the closest to training performance is “training time.” All recruits to the Metropolitan Police Department,- however, go through the *266same training course in the same amount of time, including those who experience some difficulty. See n. 7, supra. Third, the final requirement of § 1607.5 (b) (3) has not been met. There has been no job analysis establishing the significance of scores on training examinations, nor is there any other type of evidence showing that these scores are of “major or critical” importance.
Accordingly, EEOC regulations that have previously been approved by the Court set forth a construction of Title VII that is distinctly opposed to today’s statutory result.
Ill
The Court also says that its conclusion is not foreclosed by Griggs and Albemarle, but today’s result plainly conflicts with those cases. Griggs held that “[i]f an employment practice which operates to exclude Negroes cannot be shown to be related to job performance, the practice is prohibited.” 401 U. S., at 431 (emphasis added). Once a discriminatory impact is shown, the employer carries the burden of proving that the challenged practice “bear[s] a demonstrable relationship to successful performance of the jobs for which it was used.” Ibid, (emphasis added). We observed further:
“Nothing in the Act precludes the use of testing or measuring procedures; obviously they are useful. What Congress has forbidden is giving these devices and mechanisms controlling force unless they are demonstrably a reasonable measure of job performance. . . . What Congress has commanded is that any tests used must measure the person for the job and not the person in the abstract.” Id., at 436.
Albemarle read Griggs to require that a discriminatory test be validated through proof “by professionally acceptable methods” that it is “ 'predictive of or signifi*267cantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are being evaluated. ” 422 U. S., at 431 (emphasis added), quoting 29 CFR § 1607.4(c) (1975). Further, we rejected the employer's attempt to validate a written test by proving that it was related to supervisors’ job performance ratings, because there was no demonstration that the ratings accurately reflected job performance. We were unable “to determine whether the criteria actually considered were sufficiently related to the [employer’s] legitimate interest in job-specific ability to justify a testing system with a racially discriminatory impact.” 422 U. S., at 433 (emphasis in original). To me, therefore, these cases read Title VII as requiring proof of a significant relationship to job performance to establish the validity of a discriminatory test. See also McDonnell Douglas Corp. v. Green, 411 U. S. 792, 802, and n. 14 (1973). Petitioners do not maintain that there is a demonstrated correlation between Test 21 scores and job performance. Moreover, their validity study was unable to discern a significant positive relationship between training averages and job performance.10 Thus, there is no proof of a correlation — either direct or indirect — between Test 21 and performance of the job of being a police officer.
It may well be that in some circumstances, proof of a relationship between a discriminatory qualification test and training performance is an acceptable substitute for establishing a relationship to job performance. But this question is not settled, and it should not be re*268solved by the minimal analysis in the Court’s opinion.11 Moreover, it is particularly inappropriate to decide the question on this record. “Professionally acceptable methods” apparently recognize validation by proof -of a correlation with training performance, rather than job performance, if (1) the training curriculum includes information proved to be important to job performance and (2) the standard used as a measure of training performance is shown to reflect the trainees’ mastery of the material included in the training curriculum. See Brief for CSC 24-29; Brief for the Executive Committee of Division 14 of the American Psychological Assn, as Amicus Curiae 37-43. But no authority, whether professional, administrative, or judicial, has accepted the sufficiency of a correlation with training performance in the absence of such proof. For reasons that I have stated above, the record does not adequately establish either factor. As a result, the Court’s conclusion cannot be squared with the focus on job performance in Griggs and Albemarle, even if this substitute showing is reconcilable with the holdings in those cases.
Today’s reduced emphasis on a relationship to job performance is also inconsistent with clearly expressed congressional intent. A section-by-section analysis of the 1972 amendments to Title VII states as follows:
“In any area where the new law does not address itself, or in any areas where a specific contrary intention is not- indicated, it was assumed -that the present case law as developed by the courts would *269continue to govern the applicability and construction of Title VII.” 118 Cong. Rec. 7166 (1972).
The pre-1972 judicial decisions dealing with standardized tests used as job qualification requirements uniformly follow the EEOC regulations discussed above and insist upon proof of a relationship to job performance to prove that a test is job related.12 Furthermore, the Court ignores Congress’ explicit hostility toward the use of written tests as job-qualification requirements; Congress disapproved the CSC’s “use of general ability tests which are not aimed at any direct relationship to specific jobs.” H. R. Rep. No. 92-238, p. 24 (1971). See S. Rep. No. 92-415, pp. 14-15 (1971). Petitioners concede that Test 21 was devised by the CSC for general use and was not designed to be used by police departments.
Finally, it should be observed that every federal court, except the District Court in this case, presented with proof identical to that offered to validate Test 21 has reached a conclusion directly opposite to that of the *270Court today.13 Sound policy considerations support the view that, at a minimum, petitioners should have been required to prove that the police training examinations either measure job-related skills or predict job performance. Where employers try to validate written qualification tests by proving a correlation with written examinations in a training course, there is a substantial danger that people who have good verbal skills will achieve high scores on both tests due to verbal ability, rather than “job-specific ability.” As a result, employers could validate any entrance examination that measures only verbal ability by giving another written test that measures verbal ability at the end of a training course. Any contention that the resulting correlation between examination scores would be evidence that the initial test is “job related” is plainly erroneous. It seems to me, however, that the Court’s holding in this case can be read as endorsing this dubious proposition. Today’s result will prove particularly unfortunate if it is extended to govern Title VII cases.
Accordingly, accepting the Court’s assertion that it is necessary to reach the statutory issue, I would hold that petitioners have not met their burden of proof and affirm the judgment of the Court of Appeals.
ALthough I do not intend to address the constitutional questions considered by the Court in Part II of its opinion, I feel constrained to comment upon the propriety of footnote 12, ante, at 24A-245. One of the cases “disapproved” therein is presently scheduled for plenary consideration by the Court in the 1976 Term, Metropolitan *258Housing Development Corp. v. Village of Arlington Heights, 517 F. 2d 409 (CA7), cert. granted, 423 U. S. 1030 (1975). If the Court regarded this case only a few months ago as worthy of full briefing and argument, it ought not be effectively reversed merely by its inclusion in a laundry list of lower court decisions.
The only administrative authority relied on by the Court in support of its result is a regulation of the Civil Service Commission construing the civil service employment standards in Title 5 of the United States Code. Ante, at 250-251, n. 16. I note, however, that 5 U. S. C. § 3304 was brought into this case by the CSC, not by respondents, and the CSC’s only reason for referring to that provision was to establish that petitioners had been “following the job-related standards of Griggs [v. Duke Power Co., 401 U. S. 424 (1971),] for the past eighty-eight years.” Ante, at 249 n. 15.
The provision in Title YII on which petitioners place principal reliance is 42 U. S. C. § 2000e-2 (h). See Griggs v. Duke Power Co., supra, at 433-436.
See 5 CFR §300.101 et seq. (1976). These instructions contain the “regulations” that the Court finds supportive of the District Court's conclusion, which was reached under Title VII, but neither the instructions nor the general regulations are an interpretation of *260Title VII. The instructions were issued “under authority of sections 3301 and 3302 of title 5, United States Code, and E. 0, 10577, 3 CFR 1954-58 Comp., p. 218.” 37 Fed. Reg. 21552 (1972). The pertinent regulations of the CSC in 5 CFR § 300.101 et seq. were promulgated pursuant to the same authorities, as well as 5 U. S. C. §§ 7151, 7154 and Exec. Order No. 11478, 3 CFR 803 (1966-1970 Comp.).
The CSC asserts that certain of its guidelines have some bearing on Test 21’s job relatedness. Under the CSC instructions, “ ‘criterion-related’ validity,” see Douglas v. Hampton, 168 U. S. App. D. C. 62, 70 n. 60, 512 F. 2d 976, 984 n. 60 (1975), can be established by *262demonstrating a correlation between entrance examination scores and “a criterion which is legitimately based on the needs of the Federal Government.” ¶ S3 — 2 (a) (2), 37 Fed. Reg. 21558 (1972). Further, to prove validity, statistical studies must demonstrate that Test 21, “to a significant degree, measures performance or qualifications requirements which are relevant to the job or jobs for which candidates are being evaluated.” ¶83-3 (a), 37 Fed. Reg. 21558 (1972). These provisions are ignored in the Court's opinion.
On this basis, the CSC argues that the case ought to be remanded to enable petitioners to try to make such a demonstration, but this resolution seems to me inappropriate. Both lower courts recognized that petitioners had the burden of proof, and as this burden is yet unsatisfied, respondents are entitled to prevail.
The finding in the Futransky study on which the Court relies, ante, at 251 n. 17, was that Test 21 “is effective in selecting trainees who can learn the material that is taught at the Recruit School,” because it predicts averages over 85. On its face, this would appear to be an important finding, but the fact is that everyone learns the material included in the training course. The study noted that all recruits pass the training examinations; if a particular recruit has any difficulty, he is given assistance until he passes.
Still another factor mandates deference to the EEOC regulations. The House and Senate committees considering the 1972 amendments to Title VII recognized that discrimination in employment, including the use of testing devices, is a “complex and pervasive phenomenon.” S. Rep. No. 92-415, p. 5 (1971); H. R. Rep. No. 92-238, p. 8 (1971). As a result, both committees noted the need to obtain “expert assistance” in this area. S. Rep. No. 92-415, supra, at 5; H. R. Rep. No. 92-238, supra, at 8.
Indeed, two Justices asserted that the Court relied too heavily on the EEOC guidelines. 422 U. S., at 449 (Blackmun, J., concurring in judgment); id., at 451 (Burger, C. J., concurring in part and dissenting in part).
Although the validity study found that Test 21 predicted job performance for white officers, but see Albemarle, 422 U. S., at 433, no similar relationship existed for black officers. The same finding was made as to the relationship between training examination averages and job performance. See id., at 435.
The Court of Appeals recognized that deciding whether 42 U. S. C. § 2000e-2 (h) permitted such proof “is not a simple or insignificant endeavor.” 168 U. S. App. D. C. 42, 50 n. 59, 512 F. 2d 956, 964 n. 59. The court declined to express any view on this issue on the ground that petitioners had not satisfied this standard even if it were acceptable, which seems to me the proper treatment of the question.
Griggs v. Duke Power Co., 401 U. S. 424 (1971); United States v. Jacksonville Terminal Co., 451 F. 2d 418, 456-457 (GA5 1971), cert. denied, 406 U. S. 906 (1972); Hicks v. Crown Zellerbach Corp., 319 F, Supp. 314, 319-321 (ED La. 1970) (issuing preliminary injunction), 321 F. Supp. 1241, 1244 (1971) (issuing permanent injunction). See also Castro v. Beecher, 334 F. Supp. 930 (Mass. 1971), aff’d in part and rev’d in part nn other grounds, 459 F. 2d 725 (CA1 1972); Western Addition Community Org. v. Alioto, 330 F. Supp. 536, 539-540 (ND Cal. 1971), 340 F. Supp. 1351, 135A-1356 (1972) (issuing preliminary injunction), 360 F. Supp. 733 (1973) (issuing permanent injunction); Chance v. Board of Examiners, 330 F. Supp. 203 (SDNY 1971), aff’d, 458 F. 2d 1167 (CA2 1972); Baker v. Columbus Mun. Sep. School Dist., 329 F. Supp. 706, 721-722 (ND Miss. 1971), aff’d, 462 F. 2d 1112 (CA5 1972); Arrington v. Massachusetts Bay Transp. Auth., 306 F. Supp. 1355 (Mass. 1969).
United States v. City of Chicago, 385 F. Supp. 543, 555-556 (ND Ill. 1974) (police department); Officers for Justice v. CSC, 371 F. Supp. 1328, 1337 (ND Cal. 1973) (police department); Smith v. City of East Cleveland, 363 F. Supp. 1131, 1148-1149 (ND Ohio 1973) (police department), aff’d in part and rev’d in part on other grounds, 520 F. 2d 492 (CA6 1975); Harper v. Mayor of Baltimore, 359 F. Supp. 1187, 1202-1203 (Md.) (fire department), modified and aff’d, 486 F. 2d 1134 (CA4 1973); Pennsylvania v. O’Neill, 348 F. Supp. 1084, 1090-1091 (ED Pa. 1972) (police department), aff’d in pertinent part and vacated in part, 473 F. 2d 1029 (CA3 1973).