delivered the opinion of the Court.
These consolidated cases raise two important questions under Title VII of the Civil Rights Act of 1964, 78 Stat. 253, as amended by the Equal Employment Opportunity Act of 1972, 86 Stat. 103, 42 U. S. C. § 2000e et seq. (1970 ed. and Supp. Ill): First: When employees or applicants for employment have lost the opportunity to earn wages because an employer has engaged in an unlawful discriminatory employment practice, what standards should a federal district court follow in deciding whether to award or deny backpay? Second: What must an employer show to establish that pre-employment tests racially discriminatory in effect, though not in intent, are sufficiently “job related” to survive challenge under Title VII?
The respondents — plaintiffs in the District Court— are a certified class of present and former Negro employees at a paper mill in Roanoke Rapids, N. C.; the petitioners — defendants in the District Court— are the plant’s owner, the Albemarle Paper Co., and the plant employees’ labor union, Halifax Local No. 425.1 In August 1966, after filing a complaint with the Equal Employment Opportunity Commission (EEOC), and receiving notice of their right to sue,2 the *409respondents brought a class action in the United States District Court for the Eastern District of North Carolina, asking permanent injunctive relief against “any policy, practice, custom or usage” at the plant that violated Title VII. The respondents assured the court that the suit involved no claim for any monetary awards on a class basis, but in June 1970, after several years of discovery, the respondents moved to add a class demand for backpay. The court ruled that this issue would be considered at trial.
At the trial, in July and August 1971, the major issues were the plant’s seniority system, its program of employment testing, and the question of backpay. In its opinion of November 9, 1971, the court found that the petitioners had “strictly segregated” the plant’s departmental “lines of progression” prior to January 1, 1964, reserving the higher paying and more skilled lines for whites. The “racial identifiability” of whole lines of progression persisted until 1968, when the lines were reorganized under a new collective-bargaining agreement. The court found, however, that this reorganization left Negro employees “ ‘locked’ in the lower paying job classifications.” The formerly “Negro” lines of progression had been merely tacked on to the bottom of the formerly “white” lines, and promotions, demotions, and layoffs continued to be governed — where skills were “relatively equal” — by a system of “job seniority.” Because of the plant’s previous history of overt segregation, only whites had seniority in the higher job categories. Accordingly, the court ordered the petitioners to implement a system of “plantwide” seniority.
*410The court refused, however, to award backpay to the plaintiff class for losses suffered under the “job seniority” program.3 The court explained:
“In the instant case there was no evidence of bad faith non-compliance with the Act. It appears that the company as early as 1964 began active recruitment of blacks for its Maintenance Apprentice Program. Certain lines of progression were merged on its own initiative, and as judicial decisions expanded the then existing interpretations of the Act, the defendants took steps to correct the abuses without delay. . . .
“In addition, an award of back pay is an equitable remedy. . . . The plaintiffs’ claim for back pay was filed nearly five years after the institution of this action. It was not prayed for in the pleadings. Although neither party can be charged with deliberate dilatory tactics in bringing this cause to trial, it is apparent that the defendants would be substantially prejudiced by the granting of such affirmative relief. The defendants might have chosen to exercise unusual zeal in having this court determine their rights at an earlier date had they known that back pay would be at issue.”
The court also refused to enjoin or limit Albemarle’s testing program. Albemarle had required applicants for employment in the skilled lines of progression to have a high school diploma and to pass two tests, the Revised Beta Examination, allegedly a measure of nonverbal in*411telligence, and the Wonderlic Personnel Test (available in alternative Forms A and B), allegedly a measure of verbal facility. After this Court’s decision in Griggs v. Duke Power Co., 401 U. S. 424 (1971), and on the eve of trial, Albemarle engaged an industrial psychologist to study the “job relatedness” of its testing program. His study compared the test scores of current employees with supervisorial judgments of their competence in ten job groupings selected from the middle or top of the plant’s skilled lines of progression. The study showed a .statistically significant correlation with supervisorial ratings in three job groupings for the Beta Test, in seven job groupings for either Form A or Form B of the Wonderlic Test, and in two job groupings for the required battery of both the Beta and the Wonderlic Tests.4 The respondents’ experts challenged the reliability of these studies, but the court concluded:
“The personnel tests administered at the plant have undergone validation studies and have been proven to be job related. The defendants have carried the burden of proof in proving that these tests are ‘necessary for the safe and efficient operation of the business’ and are, therefore, permitted by the Act. However, the high school education requirement used in conjunction with the testing requirements is unlawful in that the personnel tests alone are adequate to measure the mental ability and reading skills required for the job classifications.”
The petitioners did not seek review of the court’s judgment, but the respondents appealed the denial of a back-pay award and the refusal to enjoin or limit Albemarle’s use of pre-employment tests. A divided Court of Appeals for the Fourth Circuit reversed the judgment of *412the District Court, ruling that backpay should have been awarded and that use of the tests should have been enjoined, 474 F. 2d 134 (1973). As for backpay, the Court of Appeals held that an award could properly be requested after the complaint was filed and that an award could not be denied merely because the employer had not acted in “bad faith,” id., at 142:
“Because of the compensatory nature of a back pay award and the strong congressional policy embodied in Title VII, a district court must exercise its discretion as to back pay in the same manner it must exercise discretion as to attorney fees under Title II of the Civil Rights Act. . . . Thus, a plaintiff or a complaining class who is successful in obtaining an injunction under Title VII of the Act should ordinarily be awarded back pay unless special circumstances would render such an award unjust. Newman v. Piggie Park Enterprises, 390 U. S. 400 ... (1968).” (Footnote omitted.)
As for the pre-employment tests, the Court of Appeals held, id., at 138, that it was error
“to approve a validation study done without job analysis, to allow Albemarle to require tests for 6 lines of progression where there has been no validation study at all, and to allow Albemarle to require a person to pass two tests for entrance into 7 lines of progression when only one of those tests was validated for that line of progression.”
In so holding the Court of Appeals “gave great deference” to the “Guidelines on Employee Selection Procedures,” 29 CFR pt. 1607, which the EEOC has issued “as a workable set of standards for employers, unions and employment agencies in determining whether their selection *413procedures conform with the obligations contained in title VII...29 CFR § 1607.1 (c).
We granted certiorari5 because of an evident Circuit conflict as to the standards governing awards of back-pay 6 and as to the showing required to establish the “job relatedness” of pre-employment tests.7
II
Whether a particular member of the plaintiff class should have been awarded any backpay and, if so, how much, are questions not involved in this review. The equities of individual cases were never reached. Though at least some of the members of the plaintiff class obviously suffered a loss of wage opportunities on account of Albemarle’s unlawfully discriminatory system of job seniority, the District Court decided that no backpay should be awarded to anyone in the class. The court declined to make such an award on two stated grounds: the lack of “evidence of bad faith non-compliance with the Act,” and the fact that “the defendants would be substantially prejudiced” by an award of backpay that was demanded contrary to an earlier representation and late in the progress of the litigation. Relying directly *414on Newman v. Piggie Park Enterprises, 390 U. S. 400 (1968), the Court of Appeals reversed, holding that back-pay could be denied only in “special circumstances.” The petitioners argue that the Court of Appeals was in error — that a district court has virtually unfettered discretion to award or deny backpay, and that there was no abuse of that discretion here.8
*415Piggie Park Enterprises, supra, is not directly in point. The Court held there that attorneys’ fees should “ordinarily” be awarded — i. in all but “special circumstances” — to plaintiffs successful in obtaining injunctions against discrimination in public accommodations, under Title II of the Civil Rights Act of 1964. While the Act appears to leave Title II fee awards to the district court’s discretion, 42 U. S. C. § 2000a-3 (b), the court determined that the great public interest in having injunctive actions brought could be vindicated only if successful plaintiffs, acting as “private attorneys general,” were awarded attorneys’ fees in all but very unusual circumstances. There is, of course, an equally strong public interest in having injunctive actions brought under Title VII, to eradicate discriminatory employment practices. But this interest can be vindicated by applying the Piggie Park standard to the attorneys’ fees provision of Title VII, 42 U. S. C. § 2000e-5 (k), see Northcross v. Memphis Board of Education, 412 U. S. 427, 428 (1973). For guidance as to the granting and denial of backpay, one must, therefore, look elsewhere.
The petitioners contend that the statutory scheme provides no guidance, beyond indicating that backpay awards are within the District Court’s discretion. We disagree. It is true that backpay is not an automatic or mandatory remedy; like all other remedies under the Act, it is one which the courts “may” invoke.9 The *416scheme implicitly recognizes that there may be cases calling for one remedy but not another, and — owing to the structure of the federal judiciary — these choices are, of course, left in the first instance to the district courts. However, such discretionary choices are not left to a court’s “inclination, but to its judgment; and its judgment is to be guided by sound legal principles.” United States v. Burr, 25 F. Cas. 30, 35 (No. 14,692d) (CC Va. 1807) (Marshall, C. J.). The power to award backpay was bestowed by Congress, as part of a complex legislative design directed at a historic evil of national proportions. A court must exercise this power “in light of the large objectives of the Act,” Hecht Co. v. Bowles, 321 U. S. 321, 331 (1944). That the court’s discretion is equitable in nature, see Curtis v. Loether, 415 U. S. 189, 197 (1974), hardly means that it is unfettered by meaningful standards or shielded from thorough appellate review. In Mitchell v. DeMario Jewelry, 361 U. S. 288, 292 (1960), this Court held, in the face of a silent statute, that district courts enjoyed the “historic power of equity” to award lost wages to workmen unlawfully discriminated *417against under § 17 of the Fair Labor Standards Act of 1938, 52 Stat. 1069, as amended, 29 U. S. C. § 217 (1958 ed.) The Court simultaneously noted that “the statutory purposes [leave] little room for the exercise of discretion not to order reimbursement.” 361 U. S., at 296.
It is true that “[e]quity eschews mechanical rules . . . [and] depends on flexibility.” Holmberg v. Armbrecht, 327 U. S. 392, 396 (1946). But when Congress invokes the Chancellor’s conscience to further transcendent legislative purposes, what is required is the principled application of standards consistent with those purposes and not “equity [which] varies like the Chancellor’s foot.” 10 Important national goals would be frustrated by a regime of discretion that “produce [d] different results for breaches of duty in situations that cannot be differentiated in policy.” Moragne v. States Marine Lines, 398 U. S. 375, 405 (1970).
The District Court’s decision must therefore be measured against the purposes which inform Title VII. As the Court observed in Griggs v. Duke Power Co., 401 U. S., at 429-430, the primary objective was a prophylactic one:
“It was to achieve equality of employment opportunities and remove barriers that have operated in the past to favor an identifiable group of white employees over other employees.”
Backpay has an obvious connection with this purpose. If employers faced only the prospect of an injunctive order, they would have little incentive to shun practices of dubious legality. It is the reasonably certain prospect of a backpay award that “provide[s] the spur or catalyst *418which causes employers and unions to self-examine and to self-evaluate their employment practices and to endeavor to eliminate, so far as possible, the last vestiges of an unfortunate and ignominious page in this country’s history.” United States v. N. L. Industries, Inc., 479 F. 2d 354, 379 (CA8 1973).
It is also the purpose of Title VII to make persons whole for injuries suffered on account of unlawful employment discrimination. This is shown by the very fact that Congress took care to arm the courts with full equitable powers. For it is the historic purpose of equity to “secur [e] complete justice,” Brown v. Swann, 10 Pet. 497, 503 (1836); see also Porter v. Warner Holding Co., 328 U. S. 395, 397-398 (1946). “[WJhere federally protected rights have been invaded, it has been the rule from the beginning that courts will be alert to adjust their remedies so as to grant the necessary relief.” Bell v. Hood, 327 U. S. 678, 684 (1946). Title VII deals with legal injuries of an economic character occasioned by racial or other antiminority discrimination. The terms “complete justice” and “necessary relief” have acquired a clear meaning in such circumstances. Where racial discrimination is concerned, “the [district] court has not merely the power but the duty to render a decree which will so far as possible eliminate the discriminatory effects of the past as well as bar like discrimination in the future.” Louisiana v. United States, 380 U. S. 145, 154 (1965). And where a legal injury is of an economic character,
“[t]he general rule is, that when a wrong has been done, and the law gives a remedy, the compensation shall be equal to the injury. The latter is the standard by which the former is to be measured. The injured party is to be placed, as near as may be, in *419the situation he would have occupied if the wrong had not been committed.” Wicker v. Hoppock, 6 Wall. 94, 99 (1867).
The “make whole” purpose of Title VII is made evident by the legislative history. The backpay provision was expressly modeled on the backpay provision of the National Labor Relations Act.11 Under that Act, “[m]aking the workers whole for losses suffered on account of an unfair labor practice is part of the vindication of the public policy which the Board. enforces.” Phelps Dodge Corp. v. NLRB, 313 U. S. 177, 197 (1941). See also Nathanson v. NLRB, 344 U. S. 25, 27 (1952); NLRB v. Rutter-Rex Mfg. Co., 396 U. S. 258, 263 (1969). We may assume that Congress was aware that the Board, *420since its inception, has awarded backpay as a matter of course — not randomly or in the exercise of a standardless discretion, and not merely where employer violations are peculiarly deliberate, egregious, or inexcusable.12 Furthermore, in passing the Equal Employment Opportunity Act of 1972, Congress considered several bills to limit the judicial power to award backpay. These limiting efforts were rejected, and the backpay provision was re-enacted substantially in its original form.13 A Section-by-Section Analysis introduced by Senator Williams to accompany the Conference Committee Report on the 1972 Act *421strongly reaffirmed the “make whole” purpose of Title VII:
“The provisions of this subsection are intended to give the courts wide discretion exercising their equitable powers to fashion the most complete relief possible. In dealing with the present section 706 (g) the courts have stressed that the scope of relief under that section of the Act is intended to make the victims of unlawful discrimination whole, and that the attainment of this objective rests not only upon the elimination of the particular unlawful employment practice complained of, but also requires that persons aggrieved by the consequences and effects of the unlawful employment practice be, so far as possible, restored to a position where they would have been were it not for the unlawful discrimination.” 118 Cong. Rec. 7168 (1972).
As this makes clear, Congress’ purpose in vesting a variety of “discretionary” powers in the courts was not to limit appellate review of trial courts, or to invite inconsistency and caprice, but rather to make possible the “fashion[ing] [of] the most complete relief possible.”
It follows that, given a finding of unlawful discrimination, backpay should be denied only for reasons which, if applied generally, would not frustrate the central statutory purposes of eradicating discrimination throughout the economy and making persons whole for injuries suffered through past discrimination.14 The courts of appeals must maintain a consistent and principled application of the backpay provision, consonant with the twin statutory objectives, while at the same time recognizing that the trial court will often have the keener apprecia*422tion of those facts and circumstances peculiar to particular cases.
The District Court’s stated grounds for denying back-pay in this case must be tested against these standards. The first ground was that Albemarle’s breach of Title VII had not been in “bad faith.”15 This is not a sufficient reason for denying backpay. Where an employer has shown bad faith — by maintaining a practice which he knew to be illegal or of highly questionable legality — he can make no claims whatsoever on the Chancellor’s conscience. But, under Title VII, the mere absence of bad faith simply opens the door to equity; it does not depress the scales in the employer’s favor. If backpay were awardable only upon a showing of bad faith, the remedy would become a punishment for moral turpitude, rather than a compensation for workers’ injuries. This would read the “make whole” purpose right out of Title VII, for a worker’s injury is no less real simply because his employer did not inflict it in “bad faith.”16 Title VII is not concerned with the employer’s “good intent or absence of discriminatory intent” for “Congress directed the thrust of the Act to the consequences of employment practices, not simply the motivation.” Griggs v. Duke *423Power Co., 401 U. S., at 432. See also Watson v. City of Memphis, 373 U. S. 526, 535 (1963); Wright v. Council of City of Emporia, 407 U. S. 451, 461-462 (1972).17 To condition the awarding of backpay on a showing of “bad faith” would be to open an enormous chasm between injunctive and backpay relief under Title VII. There is nothing on the face of the statute or in its legislative history that justifies the creation of drastic and categorical distinctions between those two remedies.18
The District Court also grounded its denial of backpay on the fact that the respondents initially disclaimed any interest in backpay, first asserting their claim five years after the complaint was filed. The court concluded that the petitioners had been “prejudiced” by this conduct. The Court of Appeals reversed on the ground “that the broad aims of Title VII require that the issue of back pay be fully developed and determined even though it was not raised until the post-trial stage of litigation,” 474 F. 2d, at 141.
*424It is true that Title VII contains no legal bar to raising backpay claims after the complaint for injunctive relief has been filed, or indeed after a trial on that complaint has been had.19 Furthermore, Fed. Rule Civ. Proc. 54 (c) directs that
“every final judgment shall grant the relief to which the party in whose favor it is rendered is entitled, even if the party has not demanded such relief in his pleadings.”
But a party may not be “entitled” to relief if its conduct of the cause has improperly and substantially prejudiced the other party. The respondents here were not merely tardy, but also inconsistent, in demanding backpay. To deny backpay because a particular cause has been prosecuted in an eccentric fashion, prejudicial to the other party, does not offend the broad purposes of Title VII. This is not to say, however, that the District Court’s ruling was necessarily correct. Whether the petitioners were in fact prejudiced, and whether the respondents’ trial conduct was excusable, are questions that will be open to review by the Court of Appeals, if the District Court, on remand, decides again to decline to make any award of backpay.20 But the standard of review will be the familiar one of whether the District Court was “clearly erroneous” in its factual findings and whether it “abused” its traditional discretion to locate “a just result” in light of the circumstances peculiar to the case, *425Langnes v. Green, 282 U. S. 531, 541 (1931). On these issues of procedural regularity and prejudice, the “broad aims of Title VII” provide no ready solution.
Ill
In Griggs v. Duke Power Co., 401 U. S. 424 (1971), this Court unanimously held that Title VII forbids the use of employment tests that are discriminatory in effect unless the employer meets “the burden of showing that any given requirement [has] ... a manifest relationship to the employment in question.” Id., at 432.21 This burden arises, of course, only after the complaining party or class has made out a prima facie case of discrimination, i. e., has shown that the tests in question select applicants for hire or promotion in a racial pattern significantly different from that of the pool of applicants. See McDonnell Douglas Corp. v. Green, 411 U. S. 792, 802 (1973). If an employer does then meet the burden of proving that its tests are “job related,” it remains open to the complaining party to show that other tests or selection devices, without a similarly undesirable racial effect, would also serve the employer’s legitimate interest in “efficient and trustworthy workmanship.” Id., at 801. Such a showing would be evidence that the employer was using its tests merely as a “pretext” for discrimination. Id., at 804-805. In the present case, however, we are concerned only with the question whether Albemarle has shown its tests to be job related.
*426The concept of job relatedness takes on meaning from the facts of the Griggs case. A power company in North Carolina had reserved its skilled jobs for whites prior to 1965. Thereafter, the company allowed Negro workers to transfer to skilled jobs, but all transferees — white and Negro — were required to attain national median scores on two tests:
“[T]he Wonderlic Personnel Test, which purports to measure general intelligence, and the Bennett Mechanical Comprehension Test. Neither was directed or intended to measure the ability to learn to perform a particular job or category of jobs. . . .
“. . . Both were adopted, as the Court of Appeals noted, without meaningful study of their relationship to job-performance ability. Rather, a vice president of the Company testified, the requirements were instituted on the Company’s judgment that they generally would improve the overall quality of the work force.” 401 U. S., at 428-431.
The Court took note of “the inadequacy of broad and general testing devices as well as the infirmity of using diplomas or degrees as fixed measures of capability,” id., at 433, and concluded:
“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.
*427Like the employer in Griggs, Albemarle uses two general ability tests, the Beta Examination, to test nonverbal intelligence, and the Wonderlic Test (Forms A and B), the purported measure of general verbal facility which was also involved in the Griggs case. Applicants for hire into various skilled lines of progression at the plant are required to score 100 on the Beta Exam and 18 on one of the Wonderlic Test’s two alternative forms.22
The question of job relatedness must be viewed in the context of the plant’s operation and the history of the testing program. The plant, which now employs about 650 persons, converts raw wood into paper products. It is organized into a number of functional departments, each with one or more distinct lines of progression, the theory being that workers can move up the line as they acquire the necessary skills. The number and structure of the lines have varied greatly over time. For many years, certain lines were themselves more skilled and paid higher wages than others, and until 1964 these skilled lines were expressly reserved for white workers. In 1968, many of the unskilled “Negro” lines were “end-tailed” onto skilled “white” lines, but it apparently remains true that at least the top jobs in certain lines require greater skills than the top jobs in other lines. In this sense, at least, it is still possible to speak of relatively skilled and relatively unskilled lines.
In the 1950’s while the plant was being modernized with new and more sophisticated equipment, the Company introduced a high school diploma requirement for entry into the skilled lines. Though the Company soon concluded that this requirement did not improve the quality of the labor force, the requirement was continued *428until the District Court enjoined its use. In the late 1950’s the Company began using the Beta Examination and the Bennett Mechanical Comprehension Test (also involved in the Griggs case) to screen applicants for entry into the skilled lines. The Bennett Test was dropped several years later, but use of the Beta Test continued.23
The Company added the Wonderlic Tests in 1963, for the skilled lines, on the theory that a certain verbal intelligence was called for by the increasing sophistication of the plant’s operations. The Company made no attempt to validate the test for job relatedness,24 and simply adopted the national “norm” score of 18 as a cut-off point for new job applicants. After 1964, when it discontinued overt segregation in the lines of progression, *429the Company allowed Negro workers to transfer to the skilled lines if they could pass the Beta and Wonderlic Tests, but few succeeded in doing so. Incumbents in the skilled lines, some of whom had been hired before adoption of the tests, were not required to pass them to retain their jobs or their promotion rights. The record shows that a number of white incumbents in high-ranking job groups could not pass the tests.25
Because departmental reorganization continued up to the point of trial, and has indeed continued since that point, the details of the testing program are less than clear from the record. The District Court found that, since 1963, the Beta and Wonderlic Tests have been used in 13 lines of progression, within eight departments. Albemarle contends that at present the tests are used in only eight lines of progression, within four departments.
Four months before this case went to trial, Albemarle engaged an expert in industrial psychology to “validate” the job relatedness of its testing program. He spent a half day at the plant and devised a “concurrent validation” study, which was conducted by plant officials, without his supervision. The expert then subjected the results to statistical analysis. The study dealt with 10 job groupings, selected from near the top of nine of the *430lines of progression.26 Jobs were grouped together solely by their proximity in the line of progression; no attempt was made to analyze jobs in terms of the particular skills they might require. All, or nearly all, employees in the selected groups participated in the study — 105 employees in all, but only four Negroes. Within each job grouping, the study compared the test scores of each employee with an independent “ranking” of the employee, relative to each of his coworkers, made by two of the employee's supervisors. The supervisors, who did not know the test scores, were asked to
“determine which ones they felt irrespective of the job that they were actually doing, but in their respective jobs, did a better job than the person they were rating against....” 27
For each job grouping, the expert computed the “Phi coefficient” of statistical correlation between the test scores and an average of the two supervisorial rankings. Consonant with professional conventions, the expert regarded as “statistically significant” any correlation that could have occurred by chance only five times, or fewer, in 100 trials.28 On the basis of these results, the District Court found that “[t]he personnel tests administered at the plant have undergone validation studies and have been proven to be job related.” Like the Court of Appeals, we are constrained to disagree.
The EEOC has issued “Guidelines” for employers seeking to determine, through professional validation studies, *431whether their employment tests are job related. 29 CFR pt. 1607. These Guidelines draw upon and make reference to professional standards of test validation established by the American Psychological Association.29 The EEOC Guidelines are not administrative “regulations” promulgated pursuant to formal procedures established by the Congress. But, as this Court has heretofore noted, they do constitute “[t]he administrative interpretation of the Act by the enforcing agency,” and consequently they are “entitled to great deference.” Griggs v. Duke Power Co., 401 U. S., at 433-434. See also Espinoza v. Farah Mfg. Co., 414 U. S. 86, 94 (1973).
The message of these Guidelines is the same as that of the Griggs case — that discriminatory tests are impermissible unless shown, by professionally acceptable methods, to be “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).
Measured against the Guidelines, Albemarle’s validation study is materially defective in several respects:
(1) Even if it had been otherwise adequate, the study would not have “validated” the Beta and Wonderlic test battery for all of the skilled lines of progression for which the two tests are, apparently, now required. The study showed significant correlations for the Beta Exam in only three of the eight lines. Though the Wonderlic Test’s Form A and Form B are in theory identical and *432interchangeable measures of verbal facility, significant correlations for one form but not for the other were obtained in four job groupings. In two job groupings neither form showed a significant correlation. Within some of the lines of progression, one form was found acceptable for some job groupings but not for others. Even if the study were otherwise reliable, this odd patchwork of results would not entitle Albemarle to impose its testing program under the Guidelines. A test may be used in jobs other than those for which it has been professionally validated only if there are “no significant differences” between the studied and unstudied jobs. 29 CFR § 1607.4 (c)(2). The study in this case involved no analysis of the attributes of, or the particular skills needed in, the studied job groups. There is accordingly no basis for concluding that “no significant differences” exist among the lines of progression, or among distinct job groupings within the studied lines of progression. Indeed, the study’s checkered results appear to compel the opposite conclusion.
(2) The study compared test scores with subjective supervisorial rankings. While they allow the use of supervisorial rankings in test validation, the Guidelines quite plainly contemplate that the rankings will be elicited with far more care than was demonstrated here.30 *433Albemarle’s supervisors were asked to rank employees by a “standard” that was extremely vague and fatally open to divergent interpretations. As previously noted, each “job grouping” contained a number of different jobs, and the supervisors were asked, in each grouping, to
“determine which ones [employees] they felt irrespective of the job that they were actually doing, but in their respective jobs, did a better job than the person they were rating against....” 31
There is no way of knowing precisely what criteria of job performance the supervisors were considering, whether each of the supervisors was considering the same criteria or whether, indeed, any of the supervisors actually applied a focused and stable body of criteria of any kind.32 There is, in short, simply no way to determine whether the criteria actually considered were sufficiently related to the Company’s legitimate interest in job-specific ability to justify a testing system with a racially discriminatory impact.
(3) The Company’s study focused, in most cases, on job groups near the top of the various lines of progression. In Griggs v. Duke Power Co., supra, the Court *434left open “the question whether testing requirements that take into account capability for the next succeeding position or related future promotion might be utilized upon a showing that such long-range requirements fulfill a genuine business need.” 401 U. S., at 432. The Guidelines take a sensible approach to this issue, and we now endorse it:
“If job progression structures and seniority provisions are so established that new employees will probably, within a reasonable period of time and in a great majority of cases, progress to a higher level, it may be considered that candidates are being evaluated for jobs at that higher level. However, where job progression is not so nearly automatic, or the time span is such that higher level jobs or employees’ potential may be expected to change in significant ways, it shall be considered that candidates are being evaluated for a job at or near the entry level.” 29 CFR § 1607.4 (c) (1).
The fact that the best of those employees working near the top of a line of progression score well on a test does not necessarily mean that that test, or some particular cutoff score on the test, is a permissible measure of the minimal qualifications of new workers entering lower level jobs. In drawing any such conclusion, detailed consideration must be given to the normal speed of promotion, to the efficacy of on-the-job training in the scheme of promotion, and to the possible use of testing as a promotion device, rather than as a screen for entry into low-level jobs. The District Court made no findings on these issues. The issues take on special importance in a case, such as this one, where incumbent employees are permitted to work at even high-level jobs without passing the company’s test battery. See 29 CFR § 1607.11.
*435(4) Albemarle’s validation study dealt only with job-experienced, white workers; but the tests themselves are given to new job applicants, who are younger, largely inexperienced, and in many instances nonwhite. The APA Standards state that it is “essential” that
“[t]he validity of a test should be determined on subjects who are at the age or in the same educational or vocational situation as the persons for whom the test is recommended in practice.” ¶ C 5.4.
The EEOC Guidelines likewise provide that “[d]ata must be generated and results separately reported for minority and nonminority groups wherever technically feasible.” 29 CFR § 1607.5 (b) (5). In the present case, such “differential validation” as to racial groups was very likely not “feasible,” because years of discrimination at the plant have insured that nearly all of the upper level employees are white. But there has been no clear showing that differential validation was not feasible for lower level jobs. More importantly, the Guidelines provide:
“If it is not technically feasible to include minority employees in validation studies conducted on the present work force, the conduct of a validation study without minority candidates does not relieve any person of his subsequent obligation for validation when inclusion of minority candidates becomes technically feasible.” 29 CFR § 1607.5 (b)(1).
“. . . [Ejvidence of satisfactory validity based on other groups will be regarded as only provisional compliance with these guidelines pending separate validation of the test for the minority group in question.” 29 CFR § 1607.5 (b)(5).
For all these reasons, we agree with the Court of Appeals that the District Court erred in concluding that *436Albemarle had proved the job relatedness of its testing program and that the respondents were consequently not entitled to equitable relief. The outright reversal by the Court of Appeals implied that an injunction should immediately issue against all use of testing at the plant. Because of the particular circumstances here, however, it appears that the more prudent course is to leave to the District Court the precise fashioning of the necessary relief in the first instance. During the appellate stages of this litigation, the plant has apparently been amending its departmental organization and the use made of its tests. The appropriate standard of proof for job relatedness has not been clarified until today. Similarly, the respondents have not until today been specifically apprised of their opportunity to present evidence that even validated tests might be a “pretext” for discrimination in light of alternative selection procedures available to the Company. We also note that the Guidelines authorize provisional use of tests, pending new validation efforts, in certain very limited circumstances. 29 CFR § 1607.9. Whether such circumstances now obtain is a matter best decided, in the first instance, by the District Court. That court will be free to take such new evidence, and to exercise such control of the Company’s use and validation of employee selection procedures, as are warranted by the circumstances and by the controlling law.
Accordingly, the judgment is vacated, and these eases are remanded to the District Court for proceedings consistent with this opinion. r, . 7 ,
7 It is so ordered.
Mr. Justice Powell took no part in the consideration or decision of these cases.[Appendix to opinion of the Court follows.]
*437APPENDIX TO OPINION OF THE COURT CHART A
Results of Validation Study
Test
N Beta W-A W-B Job Group
1. Caustic Operator, Lime Kiln Operator ................ 8 .25 1.00** .47
2. C. E. Recovery Operator, C. E. Recovery 1st Helpers & Evaporator Operators.. 12 .64** .32 .17
3. Wood Yard: Long Log Operators, Log Stackers, Small Equipment Operators & Oilers ................... 14 .00 1.00** .72*
4. Technical Services: B Mill Shift Testmen, Additive men, General Lab. Test-men, General Lab. asst., A Mill Testmen, Samplemen. .50* .75** .64*
5. B Paper Mill: Machine Tenders and Back Tenders____ 16 .00 .50** .34
6. B Paper Mill: Stock Room Operator, Stock Room 1st Helper .................. 8 -.50 .00 .00
7. B Paper Mill: 3rd Hands, 4th Hands & 5th Hands... 21 .43 .81** .60**
8. Wood Yard: Chipper Unloader, Chipper Operator, No. 2 Chain Operator..... 6 .76* -.25 1.00**
9. Pulp Mill: Stock Room Operator, Stock Room 1st Helpers ................. 8 .50 .80* .76*'
10. Power Plant: Power Plant Operator, Power Plant 1st Helper, Power Plant 2nd Helper .................. 12 .34 .75** .66*
NOTE
The job groups are identified in Chart B. N indicates the number of employees tested. A single (double) asterisk indicates the “Phi” coefficient of correlation, shown on the chart, is statistically significant at a 95% (99%) level of confidence. The other coefficients are not statistically significant.
*438CHART B
Albemarle’s Skilled Lines of Progression
Note: The numbered job groups are those examined in the validation study summarized in Chart A. Testing is no longer required for entry into the Woodyard Department.
*439
The paper mill has changed hands during this litigation, but these changes are irrelevant to the issues considered in this opinion, and the employer interest will be referred to throughout as Albemarle or the Company. The labor union is involved in only the backpay aspect of this litigation.
The relevant procedures may be found at 42 U. S. C. § 2000e-5 (f)(1) (1970 ed., Supp. III). See McDonnell Douglas Corp. v. *409Green, 411 U. S. 792, 798 (1973); Alexander v. Gardner-Denver Co., 415 U. S. 36, 44-45 (1974). See also n. 8, infra.
Under Title VII backpay liability exists only for practices occurring after the effective date of the Act, July 2, 1965, and accrues only from a date two years prior to the filing of a charge with the EEOC. See 42 U. S. C. § 2000e-5 (g) (1970 ed., Supp. Ill), Thus no award was possible with regard to the plant’s pre-1964 policy of “strict segregation.”
See infra, at 429-430.
419 U. S. 1068 (1974). The Fourth Circuit initially granted a petition to rehear this case en banc. But that petition was ultimately denied, after this Court ruled, on a certified question, that “senior circuit judges who are members of the originally assigned division hearing a case are not authorized by Congress to participate in the determination whether to rehear that case in banc.” 417 U. S. 622, 624 (1974).
For example, compare Kober v. Westinghouse Electric Corp., 480 F. 2d 240 (CA3 1973), with Pettway v. American Cast Iron Pipe Co., 494 F. 2d 211 (CA5 1974), and Head v. Timken Boiler Bearing Co., 486 F. 2d 870 (CA6 1973).
For example, compare Pettway v. American Cast Iron Pipe Co., supra, with Castro v. Beecher, 459 F. 2d 725 (CA1 1972).
The petitioners also contend that no backpay can be awarded to those unnamed parties in the plaintiff class who have not themselves filed charges with the EEOC. We reject this contention. The Courts of Appeals that have confronted the issue are unanimous in recognizing that backpay may be awarded on a class basis under Title VII without exhaustion of administrative procedures by the unnamed class members. See, e. g., Rosen v. Public Service Electric & Gas Co., 409 F. 2d 775, 780 (CA3 1969), and 477 F. 2d 90, 95-96 (CA3 1973); Robinson v. Lorillard Corp., 444 F. 2d 791, 802 (CA4 1971); United States v. Georgia Power Co., 474 F. 2d 906, 919-921 (CA5 1973); Head v. Timken Roller Bearing Co., supra, at 876; Bowe v. Colgate-Palmolive Co., 416 F. 2d 711, 719-721 (CA7 1969); United States v. N. L. Industries, Inc., 479 F. 2d 354, 378-379 (CA8 1973). The Congress plainly ratified this construction of the Act in the course of enacting the Equal Employment Opportunity Act of 1972, Pub. L. 92-261, 86 Stat. 103. The House of Representatives passed a bill, H. R. 1746, 92d Cong., 1st Sess., that would have barred, in § 3 (e), an award of backpay to any individual who “neither filed a charge [with the EEOC] nor was named in a charge or amendment thereto.” But the Senate Committee on Labor and Public Welfare recommended, instead, the re-enactment of the backpay provision without such a limitation, and cited with approval several cases holding that backpay was awardable to class members who had not personally filed, nor been named in, charges to the EEOC. S. Rep. No. 92-415, p. 27 (1971). See also 118 Cong. Rec. 4942 (1972). The Senate passed a bill without the House’s limitation, id., at 4944, and the Conference Committee adopted the Senate position. A Section-by-Section Analysis of the Conference Committee’s resolution notes that “[a] provision limiting class actions was contained in the House bill and specifically rejected by the Conference Committee,” id., at 7168, 7565. The Conference Committee bill was accepted by both Chambers. Id., at 7170, 7573.
Title 42 U. S. C. §2000e-5 (g) (1970 ed., Supp. Ill) provides:
“If the court finds that the respondent has intentionally engaged in or is intentionally engaging in an unlawful employment practice charged in the complaint, the court may enjoin the respondent from engaging in such unlawful employment practice, and order such affirmative action as may be appropriate, which may include, but is not limited to, reinstatement or hiring of employees, with or without back pay (payable by the employer, employment agency, or labor organization, as the case may be, responsible for the unlawful *416employment practice), or any other equitable relief as the court deems appropriate. Back pay liability shall not accrue from a date more than two years prior to the filing of a charge with the Commission. Interim earnings or amounts earnable with reasonable diligence by the person or persons discriminated against shall operate to reduce the back pay otherwise allowable. No order of the court shall require the admission or reinstatement of an individual as a member of a union, or the hiring, reinstatement, or promotion of an individual as an employee, or the payment to him of any back pay, if such individual was refused admission, suspended, or expelled, or was refused employment or advancement or was suspended or discharged for any reason other than discrimination on account of race, color, religion, sex, or national origin or in violation of section 2000e-3 (a) of this title.”
Eldon, L. C., in Gee v. Pritchard, 2 Swans. *403, *414, 36 Eng. Rep. 670, 674 (1818).
Section 10 (c) of the NLRA, 49 Stat. 454, as amended, 29 U. S. C. § 160 (c), provides that when the Labor Board has found that a person has committed an “unfair labor practice,” the Board “shall issue” an order “requiring such person to cease and desist from such unfair labor practice, and to take such affirmative action including reinstatement of employees with or without back pay, as will effectuate the policies of this subchapter.” The backpay provision of Title VII provides that when the court has found “an unlawful employment practice,” it “may enjoin” the practice “and order such affirmative action as may be appropriate, which may include, but is not limited to, reinstatement or hiring of employees, with or without back pay . . . .” 42 U. S. C. §2000e-5 (g) (1970 ed., Supp. III). The framers of Title VII stated that they were using the NLRA provision as a model. 110 Cong. Rec. 6549 (1964) (remarks of Sen. Humphrey); id., at 7214 (interpretative memorandum by Sens. Clark and Case). In early versions of the Title VII provision on remedies, it was stated that a court “may” issue injunctions, but “shall” order appropriate affirmative action. This anomaly was removed by Substitute Amendment No. 656, 110 Cong. Rec. 12814, 12819 (1964). The framers regarded this as merely a “minor language change,” id., at 12723-12724 (remarks of Sen. Humphrey). We can find here no intent to back away from the NLRA model or to denigrate in any way the status of backpay relief.
“The finding of an unfair labor practice and discriminatory discharge is presumptive proof that some back pay is owed by the employer,” NLRB v. Mastro Plastics Corp., 354 F. 2d 170, 178 (CA2 1965). While the backpay decision rests in the NLRB's discretion, and not with the courts, NLRB v. Rutter-Rex Mfg. Co., 396 U. S. 258, 263 (1969), the Board has from its inception pursued “a practically uniform policy with respect to these orders requiring affirmative action.” NLRB, First Annual Report 124 (1936). “[I]n all but a few cases involving discriminatory discharges, discriminatory refusals to employ or reinstate, or discriminatory demotions in violation of section 8 (3), the Board has ordered the employer to offer reinstatement to the employee discriminated against and to make whole such employee for any loss of pay that he has suffered by reason of the discrimination.” NLRB, Second Annual Report 148 (1937).
As to the unsuccessful effort to restrict class actions for back-pay, see n. 8, supra. In addition, the Senate rejected an amendment which would have required a jury trial in Title VII cases involving backpay, 118 Cong. Rec.. 4917, 4919-4920 (1972) (remarles of Sens. Ervin and Javits), and rejected a provision that would have limited backpay liability to a date two years prior to filing a complaint in court. Compare H. R. 1746, which passed the House, with the successful Conference Committee bill, analyzed at 118 Cong. Rec. 7168 (1972), which adopted a substantially more liberal limitation, i. e., a date two years prior to filing a charge with the EEOC. See 42 U. S. C. § 2000e-5 (g) (1970 ed., Supp. III).
It is necessary, therefore, that if a district court does decline to award backpay, it carefully articulate its reasons.
The District Court thought that the breach of Title VII had not been in “bad faith” because judicial decisions had only recently focused directly on the discriminatory impact of seniority systems. The court also noted that Albemarle had taken some steps to recruit black workers into one of its departments and to eliminate strict segregation through the 1968 departmental merger.
The backpay remedy of the NLRA on which the Title VII remedy was modeled, see n. 11, supra, is fully available even where the “unfair labor practice” was committed in good faith. See, e. g., NLRB v. Rutter-Rex Mfg. Co., 396 U. S., at 265; American Machinery Corp. v. NLRB, 424 F. 2d 1321, 1328-1330 (CA5 1970) ; Laidlaw Corp. v. NLRB, 414 F. 2d 99, 107 (CA7 1969).
Title VII itself recognizes a complete, but very narrow, immunity for employer conduct shown to have been undertaken “in good faith, in conformity with, and in reliance on any written interpretation or opinion of the [Equal Employment Opportunity] Commission.” 42 U. S. C. § 2000e-12 (b). It is not for the courts to upset this legislative choice to recognize only a narrowly defined “good faith” defense.
We note that some courts have denied backpay, and limited their judgments to declaratory relief, in cases where the employer discriminated on sexual grounds in reliance on state “female protective” statutes that were inconsistent with Title VII. See, e. g., Kober v. Westinghouse Electric Corp., 480 F. 2d 240 (CA3 1973); LeBlanc v. Southern Bell Telephone & Telegraph Co., 460 F. 2d 1228 (CA5 1972); Manning v. General Motors Corp., 466 F. 2d 812 (CA6 1972); Rosenfeld v. Southern Pacific Co., 444 F. 2d 1219 (CA9 1971). There is no occasion in this case to decide whether these decisions were correct. As to the effect of Title VII on state statutes inconsistent with it, see 42 U. S. C. § 2000e-7.
See Rosen v. Public Service Electric & Gas Co., 409 F. 2d, at 780 n. 20; Robinson v. Lorillard Corp., 444 F. 2d, at 802-803; United States v. Hayes International Corp., 456 F. 2d 112, 116, 121 (CA5 1972).
The District Court’s stated grounds for denying backpay were, apparently, cumulative rather than independent. The District Court may, of course, reconsider its backpay determination in light of our ruling on the “good faith” question.
In Griggs, the Court was construing 42 U. S. C. §2000e-2 (h), which provides in pertinent part that it shall not “be an unlawful employment practice for an employer to give and to act upon the results of any professionally developed ability test provided that such test, its administration or action upon the results is not designed, intended or used to discriminate because of race, color, religion, sex or national origin.”
Albemarle has informed us that it has now reduced the cut-off score to 17 on the Wonderlic Test.
While the Company contends that the Bennett and Beta Tests were “locally validated” when they were introduced, no record of this validation was made. Plant officials could recall only the barest outlines of the alleged validation. Job relatedness cannot be proved through vague and unsubstantiated hearsay.
As explained by the responsible plant official, the Wonderlic' Test was chosen in rather casual fashion:
“I had had experience with using the Wonderlic before, which is a short form Verbal Intelligence Test, and knew that it had, uh, probably more validation studies behind it than any other short form Verbal Intelligence Test. So, after consultation we decided to institute the Wonderlic, in addition to the Beta, in view of the fact that the mill had changed quite a bit and it had become exceedingly more complex in operation .... [W]e did not, uh, validate it, uh, locally, primarily, because of the, the expense of conducting such a validation, and there were some other considerations, such as, uh, we didn’t know whether we would get the co-operation of the employees that we’d need to validate it against in taking the test, and we certainly have to have that, so, we used National Norms and on my suggestion after study of the Wonderlic and Norms had been established nationally for skilled jobs, we developed a, uh, cut-off score of eighteen (18).”
In the course of a 1971 validation effort, see supra, at 411 and infra, this page and 430, test scores were accumulated for 105 incumbent employees (101 of whom were white) working in relatively high-ranking jobs. Some of these employees apparently took the tests for the first time as part of this study. The Company’s expert testified that the test cutoff scores originally used to screen these incumbents for employment or promotion “couldn’t have been . . . very high scores because some of these guys tested very low, as low as 8 in the Wonderlic test, and as low as 95 in the Beta. They couldn’t have been using very high cut-off scores or they wouldn’t have these low testing employees.”
See the charts appended to this opinion. It should be noted that testing is no longer required for some of the job groups listed.
This “standard” for the ranking was described by the plant official who oversaw the conduct of the study.
The results of the study are displayed on Chart A in the Appendix to this opinion.
American Psychological Association, Standards for Educational and Psychological Tests and Manuals (1966) (hereafter APA Standards). A volume of the same title, containing modifications, was issued in 1974. The EEOC Guidelines refer to the APA Standards at 29 CFR § 1607.5 (a). Very similar guidelines have been issued by the Secretary of Labor for the use of federal contractors. 41 CFR § 60-3.1 et seq.
The Guidelines provide, at 29 CFR §§ 1607.5 (b)(3) and (4):
"(3) 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 behaviors as revealed by careful job analyses.
“(4) In view of the possibility of bias inherent in subjective *433evaluations, supervisory rating techniques should be carefuEy developed, and the ratings should be closely examined for evidence of bias. In addition, minorities might obtain unfairly low performance criterion scores for reasons other than supervisor’s prejudice, as when, as new employees, they have had less opportunity to learn job skills. The general point is that all criteria need to be examined to insure freedom from factors which would unfairly depress the scores of minority groups.”
See n. 27, supra.
It cannot escape notice that Albemarle’s study was conducted by plant officials, without neutral, on-the-scene oversight, at a time when this litigation was about to come to trial. Studies so closely controEed by an interested party in litigation must be examined with great care.