Payne v. Travenol Laboratories, Inc.

673 F.2d 798, 28 Fair Empl. Prac. Cas. (BNA) 1212, 33 Fed. R. Serv. 2d 1582, 1982 U.S. App. LEXIS 19886, 28 Empl. Prac. Dec. (CCH) 32,647

• Court: Court of Appeals for the Fifth Circuit
• Decided: April 22, 1982
• Docket: No. 80-3764
• Status: Published

Opinions

JERRE S. WILLIAMS, Circuit Judge:

For the second time, we are asked to resolve some of the issues arising in this protracted employment discrimination class action between Willie Mae Payne and Travenol Laboratories, Inc. The district court issued a decree on August 18, 1980. The decree, while leaving unfinished the back pay segment of this case, largely concludes the formulation of injunctive relief. Both parties have appealed urging numerous errors, major and minor.¹ Having journeyed deep into the record of this case to assess the merits of both parties’ objections, we conclude that the bulk of the decree of the district court should be affirmed. We disagree with certain aspects of the decree, however, and therefore remand the case for further proceedings.

I. BACKGROUND

Nearly a decade old at the time of this appeal, this case began on March 2, 1972 when Willie Mae Payne, a black female, and several other named plaintiffs obtained right-to-sue letters and filed suit under Title VII, 42 U.S.C. § 2000e et seq. and also filed claims under 42 U.S.C. § 1981 against Travenol Laboratories. Payne attacked a battery of Travenol’s employment practices at its Cleveland, Mississippi, pharmaceutical plant as being racially discriminatory. The complaint was later amended to include charges of sex discrimination as well.

Travenol, the largest employer in Cleveland, Mississippi, manufactures and packages drugs, particularly intravenous solutions, for distribution to medical facilities. The solutions are packaged on an assembly line by employees called “operatives”. Two types of employees work on the assembly line: assemblers, who fill, cap, and inspect the bottles, and material handlers, who orchestrate and guide the flow of materials through the assembly process. Material handlers are compensated considerably better than assemblers, whose. hourly pay is only slightly above the janitorial or custodial staff. The assembler and material handler positions require no special skills or experience.

Travenol employed no blacks at its plant until prodded by the affirmative action office of the Navy in 1964 or 1965. At that point, Travenol began to hire blacks. Simultaneously, Travenol required all applicants to be screened and referred by the . Mississippi State Employment Service (MSES) and imposed a requirement that all applicants have a tenth-grade education or General Equivalency Diploma. Travenol then interviewed referred candidates and hired upon subjective appraisals of their alertness, comprehension, and cleanliness. Until 1968, Travenol had MSES refer only males for the material handler position and females for the assembler position. Travenol discontinued its reliance on MSES in 1971, but retained the tenth-grade requirement. Travenol also employs clerks, technicians, and managers. These positions have been filled both by promotions and by outside hiring. Travenol required a twelfth-grade education or General Equivalency Diploma for some of these positions and a college degree for others. Travenol’s staff above the operative level was almost exclusively white until 1974.

In 1975 the case was tried on the issue of liability, Payne representing a class of black females. The district court rendered its decision in 1976, finding that Travenol had discriminated on the basis of race and sex.² The court enjoined certain employment practices and directed the parties to submit suggestions on further relief, including constructive seniority and backpay.

On appeal, we vacated two aspects of the injunction and affirmed two others.³ First, we held that the district court’s blanket injunction against Travenol not to discriminate on the basis of race or gender in its employment practices was not specific enough to satisfy Federal Rule of Civil Procedure 65(d). Second, we held that no named plaintiff had standing to challenge Travenol’s requirement of a tenth grade education for its entry-level assembly-line positions and vacated the district court’s conclusion that the tenth-grade requirement violated Title VII. Third, we affirmed the district court’s injunction against Travenol’s use of a twelfth-grade requirement for certain positions above entry-level. Fourth, we affirmed the injunction against Travenol’s college degree requirement for other, higher-level positions.⁴

After our decision in Payne I, the district court directed that a magistrate hear the parties’ suggestions for further relief and devise a recommended order. After ruling on numerous discovery skirmishes, the magistrate prepared a report and a proposed decree which the district court adopted in full. The decree is interlocutory in that it does not provide for a back-pay award.

The decree is tailored to provide injunctive relief for the discrimination that the district court had found in its 1976 decision, taking into account our decision in 1978 and hearings held before the magistrate in 1980. Paragraph one provides that the district court will retain jurisdiction for two years to supervise the decree, provided that back-pay litigation is concluded by that date. Paragraph two reaffirms the district court’s earlier grant of injunctive relief against Travenol’s use of twelfth-grade and college-degree requirements for certain positions above entry level. Paragraph three prohibits Travenol from discriminating subjectively on the basis of race or sex in violation of Title VII or § 1981 in its hiring and promotions decisions.

Paragraphs four, five, six, seven, and nine spell out more specific actions that Travenol must take to remedy its discriminatory hiring and promotion practices. Paragraph four, for example, prohibits Travenol from establishing new facially-neutral job prerequisites except in accord with guidelines promulgated by the Equal Employment Opportunity Commission. Paragraph seven requires Travenol to post notice of job vacancies, with details about salary, shift, and other matters.

Paragraph eight provides that Travenol must write to all unhired applicant class members who had applied by February 19, 1976, and invite them to reapply for work at Travenol. Those class members who respond are to be given priority over other applicants until they have refused an offer, accepted an offer, or been rejected on legitimate grounds. Travenol is also required to sort out those applicants who were qualified for clerical work and who indicated an interest in it, and invite them to apply for clerical jobs, giving them priority over other applicants.

Paragraph ten provides for granting constructive seniority to applicant class members who were rejected for discriminatory reasons, and employee class members whose date of hire was delayed by discrimination. Paragraph eleven provides for the use of an employee class member’s constructive seniority date in recalling employees from layoff. The remaining paragraphs provide the details for administering the decree.

Both parties object to portions of the decree. First, Payne argues that the decree denies her relief because the district court erroneously excluded black males from the class she represents and also because it erroneously established March 3, 1970 as the opening date of the class. She requests that we include black males in the class and move the opening date of the class back to the period of limitations for Title VII and for § 1981. Second, Travenol argues that the plaintiffs failed to prove discrimination in hiring or promotions, and asks us to reverse any relief in the decree on these issues. Payne argues that she proved discrimination in areas in which the district court erroneously denied her relief, and asks us to broaden the decree accordingly. Third, both parties lodge various objections to the form of the decree. Travenol claims that the decree affords too much relief, if discrimination was properly found. Payne argues that the decree affords too little relief. Our appellate jurisdiction to review this interlocutory decree rests on 28 U.S.C. § 1292(a)(1). We address each issue in turn.

II. LIMITATIONS ON THE CLASS CERTIFIED BY THE DISTRICT COURT

The district court certified a class in 1972 to include all black female employees and applicants at Travenol’s plant, but limited the class to exclude all applicants who applied before March 3, 1970, and did not reapply thereafter. As initially certified, the class also included black male employees and applicants, but in 1974 the court granted the defendant’s motion to exclude black males from the class. The court made its class certification final in December, 1976, and authorized an interlocutory appeal of this order under 28 U.S.C.A. § 1292(b), but we declined to accept the appeal. The grant of injunctive relief that we review today, therefore, awards relief only to those members of the class finally certified by order of the district court in December, 1976.⁵

Payne argues that the district court denied injunctive relief to two groups of persons who were shown at trial to be victims of discrimination. These two groups are black males, and applicants who applied before March 3, 1970 but not after that date. Both groups were denied relief because they fell outside of the confines of the class certified by the district court. Payne urges us to reverse these limitations on the class in order to afford full relief for all those entitled to it.

Before we may address these arguments, however, we must decide whether our jurisdiction to hear this interlocutory appeal from the grant or denial of an injunction encompasses the power to review the definition of a class, when injunctive relief was denied to persons excluded from the class. Travenol challenges our jurisdiction, claiming that under Gardner v. Westinghouse Broadcasting Co., 437 U.S. 478, 98 S.Ct. 2451, 57 L.Ed.2d 364 (1978), class certification questions may not be heard on interlocutory appeal. Because we find that the appellees read Gardner too broadly, we reject their arguments and conclude that the definition of the class may be reviewed to the extent that it led to the denial of injunctive relief.

All present, past, and future black female employees and all present, past and future black female applicants for employment at the facility operated by the defendants at Cleveland, Mississippi, subject, however, to this limitation, that is to say, all black female applicants for employment at said facility prior to March 3, 1970, and all black female employees whose employment terminated prior to March 3, 1970, are excluded from the class; the class shall be subject to the further limitations that all black female applicants who lack a tenth grade education or equivalency or who were not referred for employment at the Cleveland facility by the Mississippi State Employment Service during the period such referral was required, are excluded from the class.

A. Jurisdiction to Review the Definition of the Class

In Gardner v. Westinghouse Broadcasting Co., supra, the plaintiff brought a Title VII suit on behalf of herself and all female applicants and employees who might be subject to sex discrimination by the defendant employer. After the district court denied a pretrial motion for class certification, the plaintiff immediately appealed under 28 U.S.C. § 1292(a)(1), arguing that the refusal to certify the class effectively denied her much of the injunctive relief she sought. Refusing to allow the appeal, the Supreme Court held that a pretrial denial of class certification was not appealable as a refusal of injunctive relief under section 1292(a)(1). The Court noted that section 1292(a)(1) creates an exception to the policy against interlocutory and fragmentary appeals. The purpose of this exception is to give litigants a chance to gain effective review of orders with irreparable consequences. Concluding that the denial of class certification at the outset of the lawsuit does not carry a threat of irreparable harm, the Court declined to place class certification questions in the category of refusals of injunctive relief that may be appealed before judgment on the merits.

We address a different question than the one faced by the Supreme Court in Gardner. Here, the district court conducted a trial on the merits, granted an injunction, and withheld portions of injunctive relief because of the definition of the class. The appellant does not seek to appeal the ruling on the composition of the class as an independent matter. Rather, she challenges the district court’s denial of injunctive relief, which, in turn, requires her to attack the order that lead to the denial. That order was the district court’s certification of the class. Our established power to review the denial of injunctive relief embraces the power to review the orders that underpin this denial.

The Gardner Court recognized that it was not presented with the question we answer today, and it limited its holding to the facts before it. In a footnote, the Court distinguished Jenkins v. Blue Cross Mutual Hospital Insurance, Inc., 538 F.2d 164 (7th Cir. 1976) (en banc), cert. denied, 429 U.S. 986, 97 S.Ct. 506, 50 L.Ed.2d 598 (1976), in which the Seventh Circuit reviewed the district court’s denial of class certification when that denial led to the refusal of a preliminary injunction. 437 U.S. at 479 n.3. This appeal fits the Jenkins mold rather than Gardner. We agree with the Jenkins court.⁶ .To hold that section 1292(a)(1) does not permit interlocutory appeals of denials of class certification, even if injunctive relief might thereby be compromised, is one thing; to hold that section 1292(a)(1) does not permit review of the district court’s definition of a class when a denial of injunctive relief turns on that ruling is quite another. We find nothing in Gardner that compels us to refrain from reviewing the contours of class composition when they shape the contours of injunctive relief.

We recently observed that otherwise unappealable orders may be reviewed when necessary to review a denial of injunctive relief. Gould v. Control Laser Corp., 650 F.2d 617, 621, n.7 (5th Cir. 1981). The Supreme Court has gone even further in authorizing review of otherwise unappealable orders in a case properly before a Court of Appeals to review a preliminary injunction ruling. In Deckert v. Independence Shares Corporation, 311 U.S. 282, 61 S.Ct. 229, 85 L.Ed. 189 (1940), the district court had denied the defendants’ motions to dismiss, then granted a preliminary injunction. The defendants appealed from the injunction and the Court of Appeals reversed not only the injunction but also the district court’s denial of the motions to dismiss. The Supreme Court held that the Court of Appeals “properly examined the interlocutory order denying the motions to dismiss, although generally it could consider such an order only on appeal from a final decision.” 311 U.S. at 287, 61 S.Ct. at 232. We have followed Deckert in Myers v. Gilman Paper Corp., 544 F.2d 837, 847 (5th Cir. 1977), cert. dismissed, 434 U.S. 801, 98 S.Ct. 28, 54 L.Ed.2d 59 (1977), modified on other grounds, 556 F.2d 758 (5th Cir. 1977), and in Mercury Motor Express, Inc. v. Brinke, 475 F.2d 1086 (5th Cir. 1973). We need not rely here on the theory of Deckert, however. The questions concerning class certification that Payne raises are directly tied to the partial denial of an injunction that is properly before us. We therefore have power to resolve the questions of class definition as an incident of our review of the injunction. Accord: Adashunas v. Negley, 626 F.2d 600, 602-03 (7th Cir. 1980).

B. Black males

We now turn to Payne’s challenge to the exclusion of black males from the class. To place this challenge in perspective we review the history of the district court’s treatment of the class as it bears on this issue. This case was filed as a class action on March 2, 1972, by three named plaintiffs: two black females, Willie Mae Payne and Alma Jean Williams, and one black male, James Williams. The complaint alleged that Travenol discriminated against all three and the class they represented on the basis of race. On November 16, 1972, the court conditionally certified the plaintiff class to include all black applicants and employees.

On May 1, 1973 the court permitted Willie Mae Payne to amend her complaint. The amended complaint alleged sex discrimination as well as race discrimination. On May 8, 1973, the court granted James Williams’ motion to withdraw from the case because of his religious views. More than one year later, on July 31, 1973, two black females, Delilah Cherry and Birdie Lee Griffin, were permitted to intervene in the case alleging race and sex discrimination. These developments prompted the defendants to move the district court to redefine the class, and on December 20, 1974, the district court did so. The new class excluded black males, and other changes were made in the composition of the class.⁷

The district court made no written findings on its reasons for excluding black males, but the transcript of a hearing on the issue reveals that the district court was swayed by two factors. First, the district court noted that no male plaintiff remained in the case after James Williams’ withdrawal. The only remaining plaintiffs were black females. Second, the court believed that a conflict existed between the interests of black males and the black female plaintiffs who sought to represent them. The black females charged sex discrimination in the assignment of material handlers. According to the court, this created a conflict because to the extent that females proved sex discrimination, the interests of males would be impaired.⁸ The court thus concluded that black females could not adequately represent the interests of black males.

After this ruling, counsel for the plaintiffs requested that black males be sent notice of their provisional exclusion from the class and of the need for one of them to step forward if black males were to remain in the case, but the district court declined. The court directed, however, that members of the new provisional class receive notice of their inclusion in the case and be given an opportunity to opt out. On December 8, 1976, after trial, the court made its class-definition order final.

The appellants contest the redefinition of the class to exclude black males on two grounds. First, Payne denies the existence of a conflict between black females and black males that warrants excluding males from the class. Second, Payne argues that black males had a stake in the litigation because of their provisional inclusion in the class, see Sosna v. Iowa, 419 U.S. 393, 95 S.Ct. 553, 42 L.Ed.2d 532 (1975), and the district court had an obligation to protect their interest. The district court, therefore, should have sent notice to black males to permit one to intervene as a plaintiff. Had this been done, the class could have been subdivided to remove the conflict between the interests of black females and black males.

The first inquiry must be whether the district court erred in denying that black females may represent a class including black males in a mixed sex and race discrimination suit. The district court’s decision on the scope of the class can be overturned only for abuse of discretion. McGowan v. Faulkner Concrete Pipe Co., 659 F.2d 554, 559 (5th Cir. 1981); Zeidman v. J. Ray McDermott & Co., 651 F.2d 1030, 1038-39 (5th Cir. 1981); Walker v. Jim Dandy Co., 638 F.2d 1330, 1334 (5th Cir. 1981).

We start with the language of Rule 23. Under Rule 23(a) one of the four prerequisites to a class action is that “the representative parties will fairly and adequately protect the interests of the class.”⁹ Fed.R.Civ.P. 23(a). The claim that black females hold interests in conflict with the interests of black males, if true, is sufficient to defeat the adequacy of their representation. “It is axiomatic that a putative representative cannot adequately protect the class if his interests are antagonistic to or in conflict with the objectives of those he purports to represent.” 7 C. Wright & A. Miller, Federal Practice & Procedure § 1768 at 638 (1972); Johnson v. Georgia Highway Express, Inc., 417 F.2d 1122, 1125 (5th Cir. 1969).

Of course, in an employment discrimination case the court must interpret Rule 23 to accommodate the substantive policies of Title VII. See Vuyanich v. Republic National Bank of Dallas, 82 F.R.D. 420 (N.D.Tex.1979). We have repeatedly upheld the maintenance of across-the-board Title VII class actions in which, for example, an employee complaining of race discrimination represents not only other employees but also applicants suffering similar discrimination.¹⁰ An across-the-board Title VII attack requires only that there be a sufficient “nexus” between the claims of the named plaintiff and the claims of the class. Payne I, 565 F.2d at 900. Yet, while Title VII policies support a broad reading of Rule 23, they do not justify the jettison of the cardinal principle that a class representative may not head a class including persons whose interests substantially conflict with his or her own. East Texas Motor Freight System, Inc. v. Rodriguez, 431 U.S. 395, 405, 97 S.Ct. 1891, 1897, 52 L.Ed.2d 453 (1977).

The conflict found by the district court here is straightforward: the female plaintiffs asserted that Travenol discriminates against females by typically assigning them to the lower-paying job of assembler while typically assigning males to the higher-paying job of material handler. The females, therefore, sought to establish that males were favored at their expense. This claim plainly draws the interests of males into conflict with the interests of females.¹¹ We are aware of no case holding that a black female plaintiff is an adequate representative of black males in a sex and race discrimination suit when the interests of the two groups conflict.¹² On the contrary, a host of district courts have refused to permit black females to represent black males in class actions alleging both race and sex discrimination when a conflict of interest appears.¹³

A court should not lightly reject a black female’s claim to represent all blacks and all females in a sex and race discrimination suit. Rather than determine the question of conflict abstractly, the court must examine the interlacing allegations of race and sex discrimination to determine whether an actual conflict exists. For example, if a black female plaintiff argues that the employer favors white males to the detriment of both females and blacks, there is no inherent obstacle to her representation of both groups. See Donaldson v. Pillsbury, 554 F.2d 825 (8th Cir. 1977), cert. denied, 434 U.S. 856, 98 S.Ct. 177, 54 L.Ed.2d 128 (1977); Vuyanich v. Republic National Bank of Dallas, supra, 82 F.R.D. at 435. In this ease, however, the district court found an actual conflict of interests and we see no error in this conclusion.

The plaintiffs argue that because both blacks and women were adequately represented at trial, the existence of a possible conflict between the two before trial is not relevant. We reject this argument. It is true that the court’s responsibility to assess the adequacy of representation is an ongoing one. The district court may decertify a class after trial if plaintiff’s trial performance showed him or her to be an inadequate class representative. Johnson v. Uncle Ben’s, Inc., 628 F.2d 419 (5th Cir. 1980), vacated and remanded on other grounds, 451 U.S. 902, 101 S.Ct. 1967, 68 L.Ed.2d 290 (1981). But the trial court is not required to hypothesize about the effect of an actual conflict of interests on the adequacy of plaintiffs’ representation. It is enough to deny representation that the conflict is actual at the outset of the trial.

The plaintiffs also argue that black males have no “legally cognizable” interest in perpetuating sex discrimination at Travenol’s plant, and we agree. But black males do have an interest in representation of their interests with undivided loyalty. The existence vel non of sex discrimination is at issue in this case. Black males are entitled to a class representative who is free from a desire to prove a claim that will impair their interests.

Finally, the plaintiffs contend that we must balance the harms to black males flowing from their inclusion in the class against the harms flowing from their exclusion. Payne maintains that the harms of excluding black males overshadow the potential harms of including them. The district court’s ruling deprives black males of the benefits accorded to class members, including constructive seniority and back pay. To include them in the class despite the conflict of interest, Payne argues, would have brought them much less harm. This argument diverts attention from the proper inquiry under Rule 23. Rule 23 forces a court to measure the adequacy of representation because of the sensitive considerations involved in binding parties not before the court to a judgment won or lost by class representatives. See Hansberry v. Lee, 311 U.S. 32, 61 S.Ct. 115, 85 L.Ed. 22 (1940). A court need not look beyond the issues in the suit before it to determine when a conflict of interest precludes adequate representation. We conclude, therefore, that the district court was within its discretion in excluding black males from the class.

Ordinarily, if a court discerns a conflict like the one in this case, the proper solution is to create subclasses of persons whose interests are in accord. Oatis v. Crown Zellerbach Corp., 398 F.2d 496, 499 (5th Cir. 1968). Of course, each subclass must be headed by a person who claims the same injury as the subclass, but who lacks the fatal conflict. See East Texas Motor Freight System, Inc. v. Rodriguez, supra; Johnson v. American Credit Co., 581 F.2d 526, 532-33 (5th Cir. 1978). In this case, after the sole black male plaintiff, James Williams, withdrew from the case, no named plaintiff existed to head a class of black males. Payne therefore argues that having provisionally included black males in the class, the district court owed black males a duty to send them notice of their pending exclusion from the case to allow one of them to step forward to intervene. We do not find this obligation in the law.

Rule 23 requires a district court to give notice to absent class members of developments in the suit in only two situations. The first is when the court certifies a class under Rule 23(b)(3) because of common questions of law or fact that predominate over other aspects of the suit and render a class action the appropriate vehicle to resolve the claims. Fed.R.Civ.P. 23(c)(2); Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 94 S.Ct. 2140, 40 L.Ed.2d 732 (1974). The second is when a class action is to be dismissed or compromised. Fed.R.Civ.P. 23(e).¹⁴ In all other cases, notice lies within the district court’s discretion. Fed.R.Civ.P. 23(d)(2). The district court thus had discretionary power to give black males notice of their impending exit from the case. Such action in general is to be encouraged.¹⁵ We cannot say, however, that the court transgressed its authority in failing to recruit a new black male plaintiff to intervene to permit subdivision of the class. Cf. United States Parole Commission v. Geraghty, 445 U.S. 388, 100 S.Ct. 1202, 1214, 63 L.Ed.2d 479 (1980) (after denying class certification, a district court must give the representative of the plaintiff class an opportunity to propose subclasses, but has no obligation to construct them itself).

Were we to accept Payne’s reasoning, a district court that provisionally certified a class but later concluded that the existing class representative was inadequate would have to send notice soliciting a new class representative. Plainly, that is not the rule in this Circuit, see Johnson v. Uncle Ben's, Inc., supra, 628 F.2d at 423, nor should it be. The rule that Payne proposes would shift a burden onto the district court that properly remains with the plaintiff.¹⁶ Only if the black males had received notice of their initial inclusion in the class, had relied on the class suit to protect their rights, and would be prejudiced as a practical matter by exclusion from the class might the district court be obligated to take some action to safeguard their interests. See Seligson v. Plum Tree, Inc., 61 F.R.D. 343, 346 (E.D.Pa.1973); Berse v. Berman, 60 F.R.D. 414, 416 (S.D.N.Y.1973). Here, there is no showing that black males relied to their detriment on the district court’s provisional inclusion of them in the class. In the absence of such a showing, we decline to hold that the district court abused its discretion in failing to give them notice. In sum, the district court on this record was justified in its discretion in believing that recruiting black males after the one black male had dropped out almost two years before would constitute the stimulation of a new law suit by the court rather than a continuation of the old.

C. Opening date for class membership

We now take up Payne’s argument that injunctive relief was improperly limited because the' district court fixed an incorrect opening date for membership in the class. The district court limited the class to black females who applied at Travenol after March 3, 1970. Neither party is certain why the district court chose this date, and the court itself gave no reason. Payne urges us to move the date back to include all applications within the limitations periods for both Title VII and Section 1981. Travenol resists any extension and maintains that despite some difficulties in the date chosen by the district court, there is no abuse of the court’s discretion. We are persuaded that the court did abuse its discretion in choosing the opening date, although our analysis differs from either of the parties’.

The opening date for membership in a class for a Title VII claim should be set by reference to the earliest charge filed by a named plaintiff. Laffey v. Northwest Airlines, Inc., 567 F.2d 429, 472 (D.C.Cir.1976) cert. denied, 434 U.S. 1086, 98 S.Ct. 1281, 55 L.Ed.2d 792 (1978); Wetzel v. Liberty Mutual Insurance Co., 508 F.2d 239, 246 (3d Cir. 1975), cert. denied, 421 U.S. 1011, 95 S.Ct. 2415, 44 L.Ed.2d 679 (1975). See also Crawford v. Western Electric Co., 614 F.2d 1300, 1309 (5th Cir. 1980). But see Chisholm v. United States Postal Service, 665 F.2d 482, 489 (4th Cir. 1981) (opening date for class membership set by reference to the date from which a named plaintiff could collect back pay). Here, that plaintiff is Willie Mae Payne, who filed her first charge on January 29,1970.¹⁷ For Title VII purposes, Payne can represent a class that includes all qualifying¹⁸ black female applicants who could have filed a timely charge with the EEOC when she did.¹⁹ See Laffey v. Northwest Airlines, Inc., supra; Wetzel v. Liberty Mutual Insurance Co., supra. The charge filing period in 1970 was ninety days after the alleged act of discrimination.²⁰ See Sanchez v. Standard Brands, Inc., 431 F.2d 455, 460 (5th Cir. 1970). Thus, the opening date for the Title VII race discrimination class, counting back ninety days from the date of the January 29 charge, is October 31, 1969. All black females who satisfy the other requirements of class membership and who applied on or after October 31, 1969, should be members of the class for purposes of Title VII race discrimination. The district court’s later date cannot be sustained.²¹

For the Title VII sex discrimination claim, the opening date of the class is October 31, 1969, the same date as for the Title VII race discrimination claim, because Payne’s January 29, 1970, charge with the EEOC alleged both race and sex discrimination. Any black female employee or applicant suffering sex discrimination on or after this date could have filed a timely charge of sex discrimination with the EEOC by the date Payne filed her charge. Thus, the sex discrimination claims take in a class of black female employees or applicants beginning on October 31, 1969.

Our inquiry is not ended here, however, because the plaintiffs alleged violations of 42 U.S.C.A. § 1981 as well as violations of Title VII. Although both of these statutes apply to employment discrimination eases, they have independent remedies and independent statutes of limitations. See Johnson v. Railway Express Agency, Inc., 421 U.S. 454, 95 S.Ct. 1716, 44 L.Ed.2d 295 (1975). Though both the district court’s decision on liability and our earlier decision in this ease failed to mention Section 1981, there can be no doubt that the Section 1981 claim was properly pleaded and adjudicated. The decree that we review today enjoins subjective employment practices in violation of Section 1981, a ruling that must be premised on a finding of Section 1981 liability. We therefore examine what the appropriate Section 1981 class opening date should be. Cf. McWilliams v. Escambia County School Board, 658 F.2d 326 (5th Cir. 1981) (the Court separately examined whether a charge had been timely filed under Title VII and whether the acts complained of occurred within the statute of limitations for Section 1981 and Section 1983 before concluding that all claims were time-barred); Crawford v. Western Electric Co., supra, 614 F.2d at 1309. Petty v. Peoples Gas Light and Coke Co., 86 F.R.D. 336, 342-43 (N.D.Ill.1979).

Statutes of limitations for Section 1981 actions ordinarily are borrowed from state law. Johnson v. Railway Express Agency, supra, 421 U.S. at 462, 95 S.Ct. at 1721. In Truvillion v. King’s Daughters Hospital, 614 F.2d 520 (5th Cir. 1980), we held that a Section 1981 claim for a discriminatory refusal-to-hire is governed by the six-year Mississippi general statute of limitations.²² 614 F.2d at 528. At least one of the named plaintiffs here, Birdie Lee Griffin, applied for a job with Travenol in 1966 after receiving an MSES referral, and was refused employment. Payne v. Travenol Laboratories, Inc., 416 F.Supp. at 252. Thus, if the proof supports it the plaintiffs may represent a Section 1981 race discrimination class that begins six years before the date when the complaint was filed in 1972, or March 2, 1966. Section 1981, of course, does not embrace sex discrimination claims, Bobo v. ITT, Continental Baking Co., 662 F.2d 340 (5th Cir. 1981).

III. PROOF OF DISCRIMINATION

We come now to the merits of the plaintiffs’ claim that Travenol’s hiring, job assignment, promotion, and pay practices were discriminatory.²³ The district court found for the plaintiffs on most but not all liability issues. On appeal, the plaintiffs charge error in the district court’s failure to find discrimination on the remaining issues. But for the educational requirements that we struck down in Payne I, 565 F.2d at 899-900, the defendant contends that the district court erred in finding any discrimination at all. We take up each issue in turn.

A. Discrimination in hiring.

At the outset, we note one difficulty in assessing the proof of discrimination in hiring. When this case was tried, the plaintiffs principally, though not exclusively, strived to prove that Travenol’s hiring of operatives discriminated against blacks because of the minimum requirement of a tenth-grade education. See Payne v. Travenol Laboratories, Inc., 416 F.Supp. at 255. Proof of such disparate impact discrimination under Title VII is governed by Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28, L.Ed.2d 158 (1971). Griggs holds that a facially neutral employment practice that imposes a greater burden on one group than another violates Title VII unless the employer establishes that the practice arises from business necessity. A disparate impact plaintiff need not show intentional discrimination. 401 U.S. at 430-32, 91 S.Ct. at 853-54. A large quantity of Payne’s evidence was aimed precisely at the Griggs elements of proof. Of course, Payne also introduced testimony and statistics undertaking to show the opportunity for and the existence of subjective, intentional discrimination in Travenol’s hiring process — disparate treatment.

Confronted with this evidence, the district court found that Travenol hired proportionately fewer blacks, stating: “This disparate result reflects the effect of a tenth-grade requirement and of the defendants’ subjective discrimination.” 416 F.Supp. at 248. The district court then struck down the tenth grade requirement, rejecting Travenol’s business necessity defense. In Payne I, we reversed the injunction against the tenth-grade requirement, holding that no named plaintiff had standing to challenge it because each had a tenth grade education during the relevant period. 565 F.2d at 898-99.²⁴

The removal of the tenth-grade requirement issue transformed this case dramatically. As Travenol properly contends, the case must now be analyzed as a disparate treatment case rather than a disparate impact case. In a disparate impact case, the plaintiff need not prove intent to discriminate, while in a disparate treatment case, such proof is crucial. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 335 n.15, 97 S.Ct. 1843, 1854, n.15, 52 L.Ed.2d 396 (1977). The plaintiffs’ proof must now be tested under disparate treatment principles.

With the removal of the tenth-grade requirement as an issue in this case, Payne’s essential challenge is that Travenol’s subjective interviewing process led to discrimination. Payne does not assert that Travenol applied distinct, facially neutral criteria to candidates in the interview that operate disproportionately to exclude blacks. Rather, Payne points to the absence of objective criteria and the resultant opportunity to discriminate subjectively in Travenol’s hiring. In support of the contention that Travenol discriminated in hiring, Payne recites the history of Travenol’s hiring, which reveals substantial pre-Act discrimination, and compares applicant flow to actual hires, stressing the consistently lower proportion of blacks than whites hired. Payne also adverts to individual instances of discrimination. This evidence falls into the classic pattern of a classwide disparate treatment case. See Hazlewood School District v. United States, 433 U.S. 299, 97 S.Ct. 2736, 53 L.Ed.2d 768 (1977); International Brotherhood of Teamsters v. United States, supra; Wilkins v. University of Houston, 654 F.2d 388, 394-95 (5th Cir. 1981); Phillips v. Joint Legislative Committee, 637 F.2d 1014, 1024-26 (5th Cir. 1981).

Travenol’s interviewing cannot be viewed as a neutral practice with a disparate effect thus subject to Griggs v. Duke Power Co., supra, as Payne argues. Hiring processes that rely heavily on subjective interviewing provide an opportunity for the intentional discrimination that lies at the heart of a disparate treatment case. Payne does not urge that Travenol has violated Title VII by seeking alert, capable operative workers; rather she charges that Travenol masked its intentional discrimination by purporting to screen applicants for these qualities while actually discriminating on the basis of race. Subjective discrimination in interviewing or in similar evaluation processes has repeatedly been held to fall into the disparate treatment branch of Title VII law. See Hazlewood School District, supra, 433 U.S. at 304, 97 S.Ct. at 2739-40; Pouncy v. Prudential Insurance Co., 668 F.2d 795, at 801 (5th Cir. 1982) (subjective evaluation of employees is not a selection procedure tested under the disparate impact model); Wilkins v. University of Houston, supra, 654 F.2d at 394-94; Phillips v. Joint Legislative Committee, supra.

Because this case now raises claims of classwide disparate treatment, we must evaluate the evidence in light of standards established by the Supreme Court in such cases in Teamsters and Hazelwood. See Pouncy v. Prudential Insurance Co., supra. Teamsters and Hazelwood establish that in a Title VII disparate treatment class action, the plaintiff has the burden of proving a prima facie case that the defendant’s regular practice was to discriminate intentionally against a protected class. The burden may be met solely with statistics if they show a sufficiently great disparity between the employer’s treatment of blacks and of whites — a disparate result. In such circumstances, statistics alone justify an inference of discriminatory motive. See Village of Arlington Heights v. Metropolitan Housing Development Corp., 429 U.S. 252, 265-66, 97 S.Ct. 555, 563-64, 50 L.Ed.2d 450 (1971). The statistical showing of disparate result may also be buttressed with evidence of a history of discrimination practiced by the employer, individual instances of discrimination, and opportunities to discriminate that exist in the employer’s decision-making processes. If the statistical disparity is insufficient alone to establish a prima facie case, the plaintiff may get over his or her initial hurdle by combining statistics with historical, individual, or circumstantial evidence. See EEOC v. American National Bank, 652 F.2d 1176 (4th Cir. 1981).

Once the plaintiff has shown a prima facie case that the employer has pursued a policy of discrimination, the employer may undertake to rebut this showing. The rebuttal can proceed along two fronts. First, proof that the plaintiff’s statistics are inaccurate or insignificant may dispel the plaintiff’s proof of disparate result. For example, the employer may show that the disparity between the percentage of minorities in his work force and in the general population results from discrimination before the enactment of Title VII. Second, the employer may seek “to provide a nondiscriminatory explanation for the apparently discriminatory result.” Teamsters, 431 U.S. at 360 n.46, 97 S.Ct. at 1867 n.46. General assertions of good faith or of hiring only the best applicants, however, are insufficient to meet this burden. Id. at 342 n.24, 97 S.Ct. at 1858 n.24. If the employer fails to rebut the plaintiff’s case, the district court may conclude that Title VII has been violated. Id. at 361, 97 S.Ct. at 1867.

Discriminatory treatment claims under § 1981 are measured by the same standards that apply to discriminatory treatment claims under Title VII. Rivera v. City of Wichita Falls, 665 F.2d 531, 534 n.4 (5th Cir. 1982); McWilliams v. Escambia County School Board, 658 F.2d 326 (5th Cir. 1981). Thus, we apply the Teamsters pattern of proof to the § 1981 claims as well as the Title VII claims in this case.

The defendants argue that because this is a disparate treatment case, the burdens of proof and production must conform , to the model established by the Supreme Court in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973), and Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981).²⁵ We have held, however, That McDonnell Douglas and Burdine establish the model of proof only for an individual disparate treatment case. In a class action we adhere to the pattern of proof set out in Teamsters and Hazelwood. Our cases have recognized the distinction between individual and class claims of disparate treatment, see Phillips v. Joint Legislative Committee, supra, 637 F.2d at 1024-27 (5th Cir. 1981), and have consistently applied the Teamsters model of proof to the class claims, e.g. Markey v. Tenneco Oil Co., 635 F.2d 497, 499 (5th Cir. 1981). Nothing in Burdine, a single-plaintiff case, suggests that the Court intended Burdine procedures to supplant the procedures for proving classwide disparate treatment announced in Teamsters and Hazelwood. See Vuyanich v. Republic National Bank of Dallas, 521 F.Supp. 656 (N.D.Tex.1981).

We now turn to the question whether the district court’s findings support an inference of classwide disparate treatment under Teamsters. In answering this question we recite the statistical and other facts reported in the opinion of the district court that can, if supported in the record, establish discrimination absent the tenth-grade requirement. Travenol asserts that the district court made a finding of discrimination only by overlaying the disparate effects of the tenth-grade education requirement and the subjective screening accomplished through interviewing. Because the district court never considered the independent effect of interviewing, Travenol maintains, there is no finding that interviewing, without regard to the tenth-grade requirement, is discriminatory. The district court’s findings, by its own admission, are “brief and to some extent conclusory.” 416 F.Supp. at 255. Nevertheless, we believe that they are adequate to support a conclusion of classwide discrimination, without regard for the tenth-grade requirement.

First, the district court found that Travenol hired proportionately fewer blacks than whites from candidates referred by MSES.²⁶ Because MSES referred only candidates possessing a tenth-grade education, this finding is untouched by our holding in Payne I that the plaintiffs could not challenge the tenth-grade requirement. The district court also found that after Travenol ceased to require referral from MSES, its hiring continued to exclude blacks disproportionately.²⁷ This finding reflects both Travenol’s imposition of the tenth-grade requirement and its subjective discrimination. The court, therefore, did not state how much of Travenol’s disproportionate exclusion of blacks was attributable to subjective discrimination, but its finding clearly implies that some of it resulted from such discrimination.

In addition to these findings, the court relied on work-force statistics.²⁸ These statistics show a slow rise in the percentage of blacks in Travenol’s plant from 1964, when there were none, to 1971, when blacks composed nearly 25% of male operatives and slightly more than 20% of the female operatives. After 1972, when the plaintiffs filed suit, the percentage of blacks grew rapidly.

Thus, the district court clearly found that Travenol’s hiring had a disparate result, even without regard to the tenth-grade requirement. Although its statistical findings are not elaborate enough to permit us to conclude that Payne established a prima facie case through statistics alone, the court also made historical, circumstantial, and individual findings that buttress the statistics. The findings of disparate result coalesce with Travenol’s history of discrimination, the opportunities in its hiring process to discriminate, and individual instances of discrimination, all to establish a prima facie case.

Travenol’s entire work force, but for custodial positions, was exclusively white until 1964.²⁹ Before 1964, Travenol hired persons, without regard to educational requirements, based on subjective impressions gained through interviewing. Community references were used heavily in the plant’s early years, but took on less importance in the 1960s. Following the passage of Title VII in 1964, the plant’s personnel manager sought the assistance of MSES in evaluating how Travenol could bring blacks into the operative work force. MSES performed a study and recommended that Travenol require a twelfth-grade education for operative positions. Travenol refused to set such a high standard, instead requiring a tenth-grade education or GED. It is unclear why Travenol decided that it needed any educational requirement for the first time only after the passage of Title VII. Travenol also invoked the services of MSES to screen and refer applicants to the plant.

Between 1965 and 1971 Travenol considered only those applicants referred by MSES. The referred applicants were then interviewed by the plant’s personnel manager. Travenol’s personnel manager between 1965 and 1967 hired applicants based on his interview impression and the recommendation of plant employees whom he consulted about the applicants. The first personnel manager initiated a practice of consulting a black employee, Julius Lucas, about black applicants, and this practice was continued by his successors until 1971. The personnel manager during 1967 and 1968 consulted members of the local community as references to supplement his reaction to applicants gained through interviewing. The personnel manager from 1968 to 1974 made his hiring decisions almost exclusively by his subjective impression at the interview.³⁰ At no time either before or after Title VII’s enactment did Travenol give its personnel managers written instructions on standards to apply in hiring.

The district court also found that all three of the named plaintiffs belonged to a racial minority, were qualified for the operative position they sought, and were rejected by Travenol. It is also abundantly clear from the district court’s opinion that Travenol continued to hire white applicants for operative positions after these rejections. Thus, each named plaintiff fulfilled the four-part test of establishing a prima facie case under McDonnell-Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973).³¹ There is no indication in the district court’s findings that Travenol rebutted the inferences of individual discrimination that arose from the plaintiffs’ showings.

These findings, evaluated in light of Teamsters, establish a strong prima facie case of disparate treatment, absent the tenth-grade requirement. The burden then fell on Travenol to rebut the plaintiffs’ showing. The district court, however, made no findings that appear to us to blunt the plaintiffs’ evidence. We conclude, therefore, that if the district court’s findings are supported by the evidence, we are compelled to hold that Travenol’s hiring violated Title VII and Section 1981.

We have now decided that the district court’s findings, if correct, warrant a conclusion that unlawful discrimination actually occurred. We now turn to evidence of discrimination in hiring to see if it supports the district court’s findings of fact. In assessing this evidence, we apply the clearly erroneous standard to the district court’s subsidiary findings of fact. Rivera v. City of Wichita Falls, 665 F.2d at 535-36. Danner v. United States Civil Service Commission, 635 F.2d 427, 430-31 (5th Cir. 1981).

Travenol asserts that the district court’s finding of a disparate result in hiring is clearly erroneous in that it is unsupported by competent statistical evidence. Travenol argues that all the statistics in the record on applicant flow combine applicants who had a tenth-grade education and applicants who were without a tenth-grade education. If Travenol were correct, the district court’s finding of a disparate result in hiring that was the product of discrimination in interviewing could not be sustained. The statistics in the record, however, amply support the district court’s findings and we may not set them aside as clearly erroneous.

First, Payne introduced statistics that show that between November 1969 and November 1971, Travenol hired only 24.3% of black females applicants referred by MSES for operative positions and 21;0% of black males, while hiring 44.3% of the white males and 51.7% of the white females.³² Payne’s evidence also shows that between 1970 and 1974, Travenol hired 44.9% of white operative-job applicants with a tenth-grade education, but only 25% of black applicants similarly qualified.³³ More specific statistics indicate that the disproportionate exclusion of blacks continued but was tapering towards the end of this period. In 1972 to 1974, Travenol hired 26.1% of the qualified white applicants while hiring 21.5% of the qualified black applicants.³⁴ The record contains no applicant flow statistics for the period between 1974 and 1976, although the decree includes class members up to February 19, 1976, the date when the district court rendered its decision on liability.

Travenol contends that flaws in Payne’s statistics render them incompetent to support the district court’s findings. We do not agree that the picture of discrimination suggested by these statistics is a mirage. While Travenol has exposed defects and gaps in the plaintiffs’ statistics, none of them are fatal.

Travenol lodges several objections to the MSES statistics for 1969-1971. First, Travenol argues that the MSES data omit three months of statistics for material handlers and nine months for assemblers. We have condemned “extravagant extrapolation” from small samples, Hester v. Southern Railway Co., 497 F.2d 1374, 1381 (5th Cir. 1974). In Hester, we refused to accept the inference that because 50% of the persons who responded to an advertisement over a two-week period were black that the same percentage would be found to have responded to a series of ads over a two-year period. Plainly, the gaps in proof alluded to here are not of the magnitude we condemned in Hester. Travenol also argues

that MSES disclaimed certainty at trial that its data was complete. Such wavering as the witnesses displayed, however, hardly warrants discarding their evidence altogether.

Second, Travenol asserts that many of the people who MSES referred to Travenol Laboratories may never have shown up for an interview. Thus, a comparison of referred applicants to actual hires omits an important variable: how many applicants Travenol failed to interview because they did not show up. Theoretically, this point is well taken, but its major premise is wholly speculative. Contrary to Travenol’s assertion, the MSES witness did not testify that “[fjailure to report [for an interview] was common.” Rather, the witness stated that MSES records failed to indicate which applicants referred for an interview actually showed up and which did not. The following colloquy then took place between the MSES witness and Travenol’s counsel.

Q: Okay. But based on your experience the probability is that there are some who were referred and did not report, is that correct?

A: Probably. I would suspect that most of these reported, but some may not have.

Thus, instead of supporting Travenol’s argument, the record indicates that there was no evidence how many applicants did not keep their interview dates, and the witness in the best position to estimate believed that most applicants did show up for interviews. Travenol also departs from the record when it asserts that the MSES witness testified that the MSES did not always keep records of hires. In fact, the testimony reveals that the opposite is true.

Travenol next objects to the applicant' flow data from 1971 to 1974. Travenol’s leading attack is that the figures are marred by duplicate applications. Rebecca Allen, the secretary for Travenol’s personnel director, testified that approximately 25% of the applications in the applicant flow charts were duplicates. Relying on Robinson v. Union Carbide Corp., 538 F.2d 652, 658 (5th Cir. 1976), Travenol contends that this level of duplicates precludes the use of applicant flow statistics.

This contention, however, is too broad. First, Robinson reached the conclusion that applicant-flow statistics were inadequate on the basis of testimony not only that duplicates existed but also that some applicants filled out three or four applications a month. This degree of repetition might well render applicant-flow statistics suspect. Rebecca Allen’s estimate that approximately 25% of the applications were duplicates, however, does not. Her earlier testimony established that only when an application expired did an applicant have to fill out a new application; otherwise the old application was renewed.³⁵ Second, and more important, Rebecca Allen specifically testified that she did not know whether black duplicates exceeded white duplicates. Thus, there was no testimony that black duplicates occurred at a greater rate than white duplicates, and we are unwilling to make such an assumption without evidence in the record to support it. Therefore, Rebecca Allen’s estimate of duplicates is not enough to render the applicant flow statistics incompetent.³⁶

Travenol also contends that comparisons between black and white rates of hire are not relevant because they fail to isolate class members: black females. Travenol relies on Jefferies v. Harris County Community Action Association, 615 F.2d 1025 (5th Cir. 1980). Jefferies does not support this ■ contention. In Jefferies we held that a black female plaintiff is entitled to prove that she suffered discrimination as a black female, even if the employer did not discriminate against either blacks as a class or females as a class. Id. at 1032. We did not imply, however, that a black female may not claim and prove that her employer discriminated against her solely because of her race or sex. In fact, the Jefferies court addressed the plaintiff’s claim that she had suffered separate racial and sexual discrimination. Thus, Travenol’s objection to Payne’s racial statistics is unsound. While Payne’s statistics are not perfect, we find that they clearly suffice to support the district court’s findings of a disparate result.

We note, however, two areas that the district court must address on remand. First, Payne argues that Travenol discriminated from 1966 to 1969, relying for statistical evidence of applicant flow solely on an affirmative action plan prepared by Travenol. The district court apparently never ruled on the admissibility of this plan, which Travenol claims is privileged. On remand, the district court must rule on the admissibility of the plan and must make findings of fact and conclusions of law on Travenol’s hiring practices between 1966 and 1969. Second, the district court awarded constructive seniority relief to all black females who applied to Travenol before February 19, 1976, but made no findings that support a conclusion that discrimination continued beyond 1974. In the absence of findings covering this two year hiatus, we cannot decide whether such relief was warranted. We therefore must remand to allow the district court to prepare findings to explain its award, or to modify the decree. See Falcon v. General Telephone Company of the Southwest, supra, 626 F.2d at 381-82.

Travenol argues that it rebutted the plaintiff’s evidence of discrimination and the district court erred in refusing to accept the rebuttal evidence. We have interpreted Teamsters’ view of rebuttal evidence to mean that:

[O]nce the plaintiff has offered some evidence of disparity, it is then the defendant’s burden to come up with more specific statistical evidence to rebut the plaintiff’s proof.

Falcon v. General Telephone Company of the Southwest, 626 F.2d at 301. Travenol offered two statistical alternatives in rebuttal, but the district court was not persuaded. We find that the district court’s rejection of Travenol’s evidence was not clearly erroneous.

First, Travenol presented statistics that it hired equal proportions of blacks and whites among applicants with their own means of transportation.³⁷ The difficulty with these statistics is that Travenol’s interviewers did not always ask applicants about their own means of transportation, particularly if the applicant received a low interview rating. Thus, Travenol cannot show that applicants it screened out in the interview lacked transportation. Moreover, the sample from which the defendant drew its conclusions excluded one-fourth of all applications for the time it covered. The defendant’s witness could not testify that the missing applications were removed randomly. Hence, a large number of applications were excluded from the defendant’s study without any showing that statistics about the remaining applications reflected the entire applicant pool. These weaknesses detract significantly from the value of the defendant’s showing.

Second, Travenol contends that its hiring of blacks equaled or exceeded the percentage of blacks among qualified persons in the population. Those population figures, according to Travenol, rebut the showing of disparate result in applicant flow statistics. We have held that an employer may repel a finding of discrimination by showing that the proportion of minorities in its work force equals or exceeds the proportion of minorities among the qualified population. Robinson v. Union Carbide Corp., supra, 538 F.2d at 658. Robinson compared general population figures to work force figures, however, only after concluding that the plaintiff’s applicant flow statistics were flawed by excessive duplicates. Id. Here, by contrast, we have concluded that the plaintiff’s applicant flow statistics survive the employer’s attack on their validity. Applicant flow figures are more revealing of an employer’s hiring practices than general population comparisons when, as here, a greater percentage of qualified blacks apply for work than qualified whites. Travenol has an obligation not to discriminate illegally among applicants who present themselves for work. Because of the heavy minority applicant flow, Travenol cannot show that it discharged this obligation by presenting evidence that the composition of its work force mirrored that of the qualified general population.

Travenol’s final line of defense against a finding of discrimination is to argue that interviewing was necessary to select alert, capable workers for its plant. Travenol produces and packages drugs, largely intravenous solutions. There is, therefore, obvious danger to human life if the drugs are contaminated or improperly labeled. Travenol argues that only through subjective interviewing could it secure qualified workers. We recognize Travenol’s interest in minimizing the risk of hiring careless employees. But to recognize this interest is not to sanction the discriminatory practices demonstrated by the plaintiffs’ evidence.

We have repeatedly held that “hiring procedures [that] rely heavily on the subjective judgments of its executives from personal interviews [create] a procedure that can easily be used to mask racially motivated hiring decisions.” Phillips v. Joint Legislative Committee, supra, 637 F.2d at 1026. This is particularly true when an all-white supervisory staff conducts the interviews, Pettway v. American Cast Iron Pipe Co., 494 F.2d 211, 240 (5th Cir. 1974); Rowe v. General Motors Corp., 457 F.2d 348, 359 (5th Cir. 1972). Travenol’s occasional consultation of a black employee does not alter the control exercised by the white personnel staff. Travenol’s assertion that interviewing was necessary amounts to no more than a claim that Travenol used interviewing to select only the best applicants. The Supreme Court rejected such a defense in Teamsters, and we similarly reject it here. Except as noted, we affirm the district court’s,award of relief for discrimination in hiring.

B. Discrimination in Initial Assignments

We now address the claim that Travenol discriminated in the initial assignment of those hired. Payne argues that the district court erred in denying injunctive relief against the disproportionate assignment of blacks to custodial jobs. The district court’s 1976 opinion makes no findings or conclusions on discrimination in this area. The magistrate’s 1980 report and recommendation for further relief rejected Payne’s request, finding no statistical foundation for it, and the district court adopted the report as its own opinion. We agree that the plaintiffs’ evidence is not sufficient.

The magistrate noted that as of November 30, 1979, blacks composed 61.9% of all operatives and 70.8% of all custodians. Believing this disparity to be insignificant, the magistrate denied injunctive relief. We do not hold that an 8.9% disparity is necessarily insignificant. We agree, however, that the disparity between the percentage of minorities in Travenol’s operative work force and the percentage in its custodial work force does not make out a prima facie case. While a concentration of minorities in the lower echelons of an organization ordinarily is probative of discrimination, here there is no showing that those assigned to the custodial work force had a tenth grade education, and were thus qualified for operative work when that requirement was in. effect. Moreover, the disparity between, the percentage of blacks in operative work and in custodial work has been steadily narrowing.³⁸ We are therefore unable to draw conclusions on Travenol’s assignment policies from bare work force statistics.

The only other statistics cited by plaintiffs show that between 1967 and 1974 Travenol assigned 6.8 of the blacks hired for entry level work to custodial jobs, while assigning only 0.7% of the whites to custodial jobs. These figures are suggestive but do not persuade us that Travenol discriminated in initial assignments of custodial workers. First, the sample is very small; Travenol averaged seven assignments to custodial work each year during the period covered by these statistics. Second, the plaintiffs admit that half of the blacks assigned to custodial work lacked a tenth-grade education and, therefore, were not eligible for operative positions. These factors considerably erode the strength of the plaintiffs’ proof on this point and the district court did not err in refusing to grant injunctive relief.

C. Discrimination in Promotions

In this section we address Payne’s contention that Travenol structured its procedure for advancing from assembler to material handler in a fashion that discriminated against females, and her contention that black females suffered discrimination in promotions above the operative level.

1. The “Hooks” Position

Payne claims that female assemblers were impeded from entry into the material handler work force by the defendant’s use of the “hooks” position as the entry job. At the time of the trial in 1976, Travenol had a material handler position that required lifting many bottles on and off the production line with plastic devices called “hooks.” The “hooks” enabled a user to lift more bottles at a time than could be lifted by hand. The hooks position was arduous; it demanded more sustained heavy lifting than any other material handler slot. When filling material handler vacancies other than the hooks, Travenol gave priority to hooks bidders even though any operative employee could bid for the position. Since hooks workers rarely declined the opportunity to transfer, “[tjhis right of first refusal ha[d] the effect of transforming any material handler vacancy into a vacancy on the ‘hooks’ jobs.” 416 F.Supp. at 262. The district court also found that females were deterred from applying for material handler positions, and those who did were forced by the hooks position to retreat back to assembler jobs much more often than men. The court therefore concluded that the use of the hooks position as the entry into material handling unlawfully discriminated against females.

In response to a Food and Drug Administration investigation in 1975 Travenol abolished the hooks position and replaced it with the “bottle stack on, bottle stack off” position. The “bottle stack” position also involves putting bottles on the line and taking them off, but is less physically taxing than the former hooks position. Travenol also ceased to give bottle stackers priority in bidding for other material handler positions. Travenol denies that the hooks position did discriminate against women, and maintains that in any event the new bottle stack position is not sexually discriminatory. We affirm the district court’s conclusion that the hooks position discriminated against females, but agree with Travenol that its modifications of this position remove its discriminatory character.³⁹

The plaintiffs presented statistics that more females leave the bottle stack position than men. The plaintiffs also presented testimony of two females who were loath to accept bottle stack positions because they require much more physical effort than other material handler positions. While this evidence indicates that bottle stacking is still undesirable, we have no warrant for overturning the magistrate’s judgment, approved by the district court, that no injunctive relief against it is warranted. Because females no longer must become bottle stackers in order to ascend to other material handler positions, the discriminatory effect created by the hooks position is not present.

2. Promotions Above the Operative Level

The district court found Travenol guilty of a variety of discriminatory promotion practices, and the decree we review today contains provisions designed to enjoin this discrimination. Travenol argues that the evidence failed to establish the existence of racial discrimination in promotions, and, therefore, the relief directed to this issue should be struck down.

The district court relied on two major categories of evidence in concluding that Travenol discriminated in promotions. First, the district court cited statistics that leave no doubt that Travenol’s work force above the operative level was virtually all white through 1974.⁴⁰ Second, the district court examined Travenol’s promotion procedures and found them marred by the same opportunity for subjective discrimination that existed in the hiring process. Travenol attacks both findings on appeal.

Travenol argues that work force statistics are not properly considered in assessing discrimination in promotions because we determined in Payne I that many positions above the operative level were filled in part by lateral hiring. 565 F.2d at 895. In Payne I, we held that the district court did not err in measuring the disparate impact of Travenol’s twelfth-grade requirement for positions above the operative level by reference to general population figures. We noted, however, that “the large majority of the employees occupying positions subject to the twelfth grade requirement were promoted from operative jobs. . . .” Id. Thus, it is plain that work-force statistics are relevant to the impact of Travenol’s promotion procedures. See Johnson v. Uncle Ben’s, Inc., supra, 628 F.2d at 425 (when positions are filled both by lateral hiring and by promotion, “[a]s a general matter . . . cases are dealt with in terms of the extreme [exclusive promotion or exclusive lateral hiring] to which they most closely accord”).

Travenol also attempts to rebut the work-force figures by presenting statistics on actual promotions. Travenol’s evidence is that 23.2% of all promotions between 1970 and 1974 went to blacks, and during this period blacks averaged 29.9% of all operatives. The district court apparently rejected these rebuttal statistics, and the record supports this evidentiary choice. The testimony of the defendant’s witness shows that the charts from which Travenol drew its statistics are flawed by a failure clearly to categorize the positions filled through promotions, and by omissions in entries. These errors detract from the force of the defendant’s proof, and the district court’s decision to reject it is not clearly erroneous.

Travenol next argues that the process through which it selected operatives for promotions was free from bias. The district court’s finding that Travenol did have the opportunity to discriminate in promotions, however, is well supported by the evidence and by our cases. The district court found that Travenol’s failure to post notice of openings, and its reliance on the recommendations of supervisors, the length of an employee’s service, and educational requirements in making promotion decisions all operated discriminatorily. We affirmed the district court’s conclusion that the educational requirements were discriminatory in Payne I, and we do not further address them here.

Travenol argues that the failure to post notice of job vacancies is irrelevant because it did not consider requests for promotions to supervisory positions, and did not require applications for promotions to office, clerical or technical positions. The district court concluded, however, that job vacancy exposure was greater among whites than blacks. Even if applications were not required for certain positions, when most of the supervisors are white, failure to post job vacancies can only discourage applications by blacks to a greater degree than it discourages whites, who garner information through the grapevine. See e.g. Long v. Sapp, 502 F.2d 34, 41 (5th Cir. 1974).

The district court found that a major factor in promotions, the supervisor’s recommendation, was delivered without guidance by any objective standards and operated to discriminate. “This court has recognized that promotion systems utilizing subjective evaluations by all white supervisors provide a ready mechanism for discrimination.” Fisher v. Proctor & Gamble Manufacturing Co., 613 F.2d 527, 544 (5th Cir. 1980); James v. Stockham Valves & Fittings Co., 559 F.2d 310, 330 (5th Cir. 1977). We uphold this finding of the district court.

The district court also found that Travenol’s consideration of length of service as a criterion for advancement tended to perpetuate past discrimination in the hiring of operatives. Travenol argues that under the Supreme Court’s decision in Teamsters, supra, the district court improperly premised Title VII liability on the length-of-service factor. Teamsters holds that a bona fide seniority system does not violate Title VII merely because it perpetuates the effects of discrimination before the enactment of Title VII. 433 U.S. at 352-54, 97 S.Ct. at 1863-64. We see no application of Teamsters to the instant case. Travenol’s informal and erratic reliance on length of service falls far short of being a bona fide seniority system under the statute and under Teamsters.

The district court’s finding of discrimination in promotions must be affirmed.

D. Discrimination in Merit System Pay

For professional, technical, clerical, and supervisory positions, Travenol determines salary subjectively according to its merit pay system. Payne argued below that Travenol discriminated against both blacks and women in setting merit pay salaries. The district court did not address this point in its 1976 opinion, but implicitly rejected the plaintiffs’ claim when declining to accept proposed findings of fact and conclusions of law with respect to it in 1976. Payne now asks us to reverse the district court’s implicit exoneration of Travenol on the merit pay issue.

Ordinarily we would be reluctant to decide an issue on which the district court was silent. In this instance, however, we affirm the district court’s refusal to find discrimination in merit pay. Payne has called no statistics to our attention upon which a finding of discrimination can be based. Payne compares several sets of statistics that show disparities between the pay of blacks and whites or between males and females. Each comparison, however, looks at groups that differ in significant ways in addition to race or sex. For instance, Payne asserts that among employees with a twelfth-grade education, white males make $2,757.11 per year more than black females. Payne does not tell us, however, how many years the white males have worked compared to the black females. Nor does she compare white males to black females who occupy the same job category. Yet length of service and job category could clearly account for the pay differential without regard to race or sex. Finally, the comparison of black females and white males includes all job categories at the plant, not only merit pay jobs. The average salary of all jobs tells us little about inequities, if they exist, in merit pay jobs.

Similar criticisms can be leveled at every example of disparities in merit pay that Payne alleges.⁴¹ We need not engage in a sophisticated analysis to conclude that Payne has failed to isolate the influence of sex or race on merit pay at Travenol. Thus, the showing is insufficient to entitle her to relief and we affirm the district court’s denial of it.

III. CHALLENGES TO THE RELIEF ORDERED

Payne raises a number of challenges to the relief ordered in the decree. We have considered each of these objections and conclude that in most aspects, the district court did not abuse its discretion in shaping relief. Two objections with respect to constructive seniority, however, have merit.

Paragraph 10(a) of the decree provides that notice shall be given to each black woman currently working in an hourly job at Travenol “who applied for employment prior to February 19, 1976, and was not hired by that date.”⁴² Payne correctly urges that a class member whose date of hire was delayed by discrimination but who was eventually hired before February 19, 1976 is also entitled to constructive seniority. The decree must allow a black female operative who applied to Travenol during the time that unlawful employment practices prevailed at Travenol’s plant to make a prima facie showing that she would have been hired earlier but for discrimination, even if she was in fact hired before the date when Travenol ceased to violate Title VII. Such a showing is made if the time between her application and date of hire is longer than the average waiting time for white applicants who applied the same time she did.

Paragraph 11 of the decree provides that employees laid off and awaiting recall shall be recalled by seniority date, whether actual or constructive.⁴³ The use of constructive seniority, we find, must go further than the decree provides to afford a complete remedy. Under the decree an employee with normal seniority could fare better than a class member with equal constructive seniority if the employee with normal seniority has already been recalled after a layoff while the employee with constructive seniority still awaits recall.⁴⁴ In fact, an employee with normal seniority may have escaped lay off altogether while a class member with less normal seniority (because of a discriminatory delay in hire) but equal constructive seniority has been laid off. The decree fails to make whole a class member in either situation. The decree therefore must be modified on remand to allow a class member in either position to be compensated for lost earnings.

Travenol also challenges two paragraphs of the decree as to the nature of the relief granted. Paragraph 10(b) provides for the establishment of the constructive seniority date of an employee class member who claims that her date of hire was discriminatorily delayed beyond February 19, 1976. Travenol argues that paragraph 10 denies it the right to contest a constructive seniority date by proving that the delay in hire was not discriminatory.

The parties agree that an employer must have the right to prove that an employee is not entitled to constructive seniority because her delay in hire was not caused by discrimination; they disagree whether Paragraph 10 affords Travenol that right. The decree now establishes a means of calculating the presumptive constructive seniority date, then allows Travenol to suggest an alternative date. To remove any ambiguity, the district court on remand should modify paragraph 10 to provide explicitly that Travenol may dispute a class member’s entitlement to any constructive seniority at all as well as to suggest a date different from the presumptive date.

Paragraph 8 provides that Travenol must write to all unhired members of the plaintiff class of whom they know, invite them to reapply for a position, and give them priority over other applicants. Travenol contends that Paragraph 8 gives priority to all applicants, even those rejected for nondiscriminatory reasons, and, therefore, grants the plaintiff class greater relief than it deserves. Travenol also argues that not all victims of discrimination should be given hiring priority because Travenol could not have hired all of them even if it hired only black females. Travenol must be afforded the right to contest priority by arguing that an applicant was rejected for legitimate reasons. In such disputes each applicant class member is entitled to a presumption that she would have been hired but for discrimination. The burden will be on Travenol to prove that it rejected her for legitimate nondiscriminatory reasons. See Teamsters, supra, 431 U.S. at 360-62, 97 S.Ct. at 1867-68.

Travenol’s additional objection to paragraph 8, that it grants priority to too many applicants-discriminatees, is premature. The Supreme Court noted in Teamsters that “after the victims have been identified and their rightful place determined, the district court will again be faced with the delicate task of adjusting the remedial interest of discriminatees and the legitimate expectations of other employees innocent of any, wrongdoing.” 431 U.S. at 372, 97 S.Ct. at 1873. The district court here will be faced with the delicate task of determining to what extent the applicant class members will be entitled to priority over innocent current applicants.

Before the district court balances these interests, the number of applicant class members who now seek work at Travenol must be determined. It is not for us to place a ceiling on applicant class members entitled to priority. The district court must exercise its discretion in this matter after the facts have been developed more fully. We note only that the number of applicants entitled to priority need not correspond exactly to the number of black females that Travenol would have hired absent discrimination.

IV. ATTORNEY’S FEES

Payne requests that we direct an award of interim attorney’s fees. See James v. Stockham Valves & Fittings Co., 559 F.2d 310, 358-59 (5th Cir. 1977). The record reveals that the district court entered an interim fee award on December 8, 1976. This order is not now before us. Requests for further interim attorney’s fees should be addressed in the first instance to the district court.

V. CONCLUSION

In summary, we hold:

I. Our jurisdiction to review this Title VII interlocutory decree includes the power to review limitations on the scope of the class when those limits have been the basis for granting or withholding injunctive relief.

II. The district court did not abuse its discretion in redefining the class to exclude black males once the sole black male plaintiff dropped out of the case, nor was it obligated to give black males notice of the redefinition.

III. The district court erred in setting the opening date of the class. For the Title VII race and sex claims, the class should open on October 21, 1969 except as to any claims based upon the tenth-grade education requirement. For Section 1981, the class may open on March 2, 1966, if the district court finds racial discrimination going back to that date.

IV. We affirm the findings of discrimination: (a) in the hiring of black women, and (b) in promotions, to the extent that these findings support challenged portions of the decree.

V. We remand for the district court to prepare findings and conclusions on discrimination in hiring on a racial basis from 1966 to 1969 and from 1975 to February 19, 1976 on the basis of race or sex.

VI. We affirm the denial of injunctive relief against discrimination: (a) in the initial assignment of custodians, (b) in the “bottle stack on, bottle stack off” position, and (c) in the merit pay system.

VII. We affirm the other parts of the decree, except in four particulars. First, paragraph 10(a) must provide for an opportunity for black females who were hired before Travenol ceased to discriminate, but whose date of hire was discriminatorily delayed, to claim constructive seniority. Second, paragraph 11 must be modified to compensate class members for lost earnings when layoffs have fallen more heavily on them than on employees who were not the subject of discrimination by Travenol. Third, paragraph 8 must be modified to allow Travenol to prove that applicant class members were rejected for legitimate reasons. Finally, paragraph 10(b) must be modified to allow Travenol to prove that an operative class member’s date of hire was not delayed because of discrimination.

AFFIRMED IN PART, REVERSED IN PART, AND REMANDED.

1 . In this opinion we shall refer to Payne as the plaintiff or the appellant, and to Travenol as the defendant or the appellee.

2 . The district court’s opinion on Travenol’s liability is reported in Payne v. Travenol Laboratories, Inc., 416 F.Supp. 248 (N.D.Miss.1976), affirmed in part and reversed and vacated in part, 565 F.2d 895 (5th Cir. 1978), cert. denied, 439 U.S. 835, 99 S.Ct. 118, 58 L.Ed.2d 131 (1978).

3 . 565 F.2d 895 (5th Cir. 1978), cert. denied, 439 U.S. 835, 99 S.Ct. 118, 58 L.Ed.2d 131 (1978).

4 . Both parties agreed that a paragraph of the injunction concerning disability coverage for pregnancy had to be vacated in the light of General Electric Co. v. Gilbert, 429 U.S. 125, 97 S.Ct. 401, 50 L.Ed.2d 343 (1976). See Payne I, 565 F.2d at 897.

5 . The district court certified the following class:

6 . The case for reviewing the class question here is even stronger than in Jenkins. In Jenkins, the district court denied a pretrial motion to certify a class, and then refused a preliminary injunction. The Court of Appeals entertained an appeal of this ruling before a trial on the merits as is authorized by section 1291(a)(1). Here, by contrast, though we are not reviewing a final judgment, we are presented with a case in a considerably advanced stage of litigation. A two week trial on the merits has occurred, as well as further hearings aimed at developing appropriate relief; the class question has been treated as final by the district court for at least five years; and the decree entered by the court is a comprehensive resolution of all issues in the case except back pay. Gardner sought to promote the policies against fragmentary and uncoordinated interlocutory appeals. These policies have much less force when applied to a case like this one. Judicial economy is furthered by our present consideration of the class definition issues. To postpone consideration of these issues until after the adjudication of back pay only threatens to delay final resolution of this case.

7 . The plaintiffs had sought in the trial court to represent white females. The district court denied the motion on December 20, 1974, and the plaintiffs do not challenge that ruling on this appeal.

8 . The district court stated: “1 think there is too much conflict between the males and the females in this situation for them to represent a class which would involve black males, or males of either race. There is a controversy in this case with reference to whether or not females should be permitted to be employed in the position of material handler in the plant, because they have always been assigned to and employed in the assembly part of the plant, and in that category which carries, as I understand it, a lower rate of pay. If these black female plaintiffs are going to get into a controversy with males, either white or black, then I don’t think they can represent the male population of the plant so far as males or [sic] concerned, or for that matter any males who may have any interests in retaining their jobs as material handlers.”

9 . All four prerequisites of Rule 23(a) must be met before certification of a class is appropriate. Huff v. N. D. Cass Co., 485 F.2d 710 (5th Cir. 1973) (en banc). Of course, the court must go on to decide if the further requirements of Rule 23(b) are satisfied before certifying the class.

10 . See Phillips v. Joint Legislative Committee, 637 F.2d 1014 (5th Cir. 1981); Falcon v. General Telephone Company of the Southwest, 626 F.2d 369 (5th Cir. 1980), vacated and remanded on other grounds, 450 U.S. 1036, 101 S.Ct. 1752, 68 L.Ed.2d 234 (1981), reinstated in part and remanded in part, 647 F.2d 633 (1981), cert. granted, --- U.S. ---, 102 S.Ct. 668, 70 L.Ed.2d 637 (1981); Long v. Sapp, 502 F.2d 34 (5th Cir. 1974), Carr v. Conoco Plastics, Inc., 423 F.2d 57 (5th Cir. 1970), cert. denied, 400 U.S. 951) 91 S.Ct. 241, 27 L.Ed.2d 257 (1970), Johnson v. Georgia Highway Express, supra. We upheld the right of the plaintiffs in this case to prosecute an across-the-board Title VII action in Payne I, 567 F.2d at 900.

11 . "On the one hand, [the black female plaintiff] would argue that defendant’s policies and practices discriminated in favor of males, including black males, and against females. At the same time, she would claim that defendant favored whites, including white females, over blacks. In light of these conflicting positions, plaintiff would not be an adequate representative of either black males or white females.” Colston v. Maryland Cup Corp., 26 F.R.Serv.2d 940, 943 (D.Md.1978).

12 . The only Circuit to have ruled on the propriety of a black female representing blacks and females in a mixed sex and race discrimination suit is the Eighth. In Donaldson v. Pillsbury, 554 F.2d 825 (8th Cir. 1977), cert. denied, 434 U.S. 856, 98 S.Ct. 177, 54 L.Ed.2d 128 (1977), the Court held that a black female alleging both race and sex discrimination had claims that were typical of all female and all black employees and, therefore, she could represent them. “Typicality” under Rule 23(a)(3) was the sole issue in the case; the Court did not discuss possible conflicts of interest between the plaintiff and the class that might render her representation inadequate under Rule 23(a)(4). Moreover, the plaintiff in Donaldson alleged that the defendant employer favored white men at the expense of both blacks and females. 554 F.2d at 830. Thus, the conflict we find here was not present in Donaldson.

13 . See Strong v. Arkansas Blue Cross & Blue Shield, Inc., 87 F.R.D. 496 (E.D.Ark.1980); Hammons v. Folger Coffee, Inc., 87 F.R.D. 600 (W.D.Mo.1980); Edmondson v. Simon, 86 F.R.D. 375 (N.D.Ill.1980); Vuyanich v. Republic National Bank of Dallas, 82 F.R.D. 420 (N.D.Tex.1979); Colston v. Maryland Cup Corp., 26 F.R.Serv.2d 940 (D.Md.1978); Droughn v. FMC Corp., 74 F.R.D. 639 (E.D.Pa.1977); Martinez v. Bechtel Corp., 21 F.R.Serv.2d 85 (N.D.Cal.1975); Arey v. Providence Hospital, 55 F.R.D. 62 (D.D.C.1972).

14 . The plaintiffs have not argued that the redefinition of the class in this case is a dismissal within the meaning of Rule 23(e), and we do not consider it to be one.

15 . In Silva v. Voweli, 621 F.2d 640, 649 (5th Cir. 1980) (quoting Johnson v. American Credit Co., supra, 581 F.2d at 533 n.13), cert. denied, 449 U.S. 1125, 101 S.Ct. 941, 67 L.Ed.2d 111 (1981), we stated that if the court had found no named plaintiff qualified to represent a subclass, “the court could have considered ‘whether it is in the interest of justice and judicial economy to postpone dismissal as to the subclass for a specified period in which members of the subclass could become plaintiffs by amendment of the Complaint or by intervention and thereby save the subclass action.’ ” In Sullivan v. Winn-Dixie Greenville Inc., 62 F.R.D. 370, 377 (D.S.C.1974), the court followed this procedure but did not require notice to the subclass members. But see Alexander v. Avco Corp., 380 F.Supp. 1282, 1286 (N.D.Tenn. 1974), (the court gave notice to class members to give them an opportunity to intervene as individuals before dismissing the class action) modified, 565 F.2d 1364 (6th Cir. 1977), cert. denied, 436 U.S. 946, 98 S.Ct. 2849, 56 L.Ed.2d 787 (1978).

16 . At oral argument, plaintiffs’ counsel advised this Court that counsel failed to seek a black male plaintiff to intervene on his own because of doubts that attorneys for the plaintiffs could freely communicate with class members. Gulf Oil Co. v. Bernard, 452 U.S. 89, 101 S.Ct. 2193, 68 L.Ed.2d 693 (1981) made clear that a district court may not curtail communications between the attorney for the plaintiff and class members, at least in the absence of specific findings that such an order is appropriate or necessary to prevent abuses. 101 S.Ct. at 2193. Although in 1974 the plaintiffs lacked the guidance of Gulf Oil, nothing in the record indicates that the plaintiffs asked the district court to allow them to communicate with the class. We cannot impose a duty on the district court to fill the void created by the plaintiffs’ failure to press their rights to communicate with persons excluded from the class.

17 . There are at present three named plaintiffs in this case. Willie Mae Payne acquired a tenth grade GED on January 20, 1970. She applied at the Mississippi State Employment Service (MSES) for a referral to Travenol on January 29, 1970, but MSES did not refer her. Payne went to Travenol anyway, applied, and was refused a job. She filed two charges with the EEOC, the first on January 29, 1970 because of Travenol’s rejection of her on that date, then on March 3, 1970, because of Travenol’s rejection of her again on February 24. On March 17, 1970, Payne was referred by MSES and was refused employment a third time. No charge appears to have been filed in reference to the March 17 application. Nothing in the record indicates that MSES referred Payne before March 17.

Delilah Cherry obtained her GED in 1967, was referred by MSES shortly thereafter, and was denied employment over the next two years despite repeated visits and phone calls to Travenol. She filed two EEOC charges, the first on August 14, 1970, the second on January 12, 1971.

Birdie Lee Griffin completed eleven years of school, was referred by MSES in 1966, and was refused employment despite renewed requests on abundant instances over the next four years. Her first EEOC charge is dated August 28, 1970. Like the other two plaintiffs, she filed a second charge, which the EEOC received on January 12, 1971.

18 . The district court certified a class of black female applicants subject to two limitations, which are in addition to the date of application. First, the applicant must possess a tenth grade education. Second, she must have been referred by MSES during the time such referral was required.

19 . Payne’s first charge covers Travenol’s alleged discriminatory acts that occurred after she filed it because they are sufficiently related to the allegations of the first charge. Gamble v. Birmingham Southern Railroad Co., 514 F.2d 678, 689 n.6 (5th Cir. 1975); Oubichon v. North American Rockwell Corp., 482 F.2d 569 (9th Cir. 1973). See generally Sanchez v. Standard Brands, Inc., supra. Thus, the January 29, 1970 charge encompasses Travenol’s rejection of Payne on March 17, 1970, after she was referred by the MSES.

20 . The statute was amended in 1972 to provide for a 180-day charge filing period. See Equal Employment Opportunity Act of 1972, Pub. L.No. 92-261, 86 Stat. 113, codified at 42 U.S. C.A. § 2000e.

We hold that the ninety-day charge-filing period, which was in effect before the 1972 amendments to Title VII, should govern this case. We reach this conclusion even though the right-to-sue letter on the charge that we use to set the opening date of the class was not issued until September 6, 1972, after the effective date of the 1972 amendments on March 24, 1972. Because Payne’s amended complaint relates back to the date of the original complaint, Fed.R.Civ.P. 15(c), this case clearly was not pending with the EEOC on the effective date of the 1972 amendments. Therefore, the ninety day period is binding on Payne and the class she represents.

21 . The court apparently chose this date because it was the date of Payne’s second EEOC charge. When the original complaint in federal court was filed on March 2, 1972, Payne did not allege having filed charges with the EEOC on January 29,1970; only the March 3 charge was mentioned in the complaint. The district court certified the class conditionally on November 16, 1972. Payne’s amended complaint, filed May 3, 1973, did allege having filed a chárge with the EEOC on January 29, 1970, but the district court failed to modify the opening date of the class to reflect the amended complaint. In any event, the court should have set the opening date of the class to include all persons who could have filed charges when Payne did instead of setting the opening date of the class on the date when she filed her charge.

22 . Mississippi Code Annotated, Section 15-1-49 (1972) provides:

All actions for which no other period of limitations is prescribed shall be commenced within six years next after the cause of such action accrued, and not after.

23 . The plaintiffs alleged discrimination in four areas that are relevant to this appeal:

1. Discrimination in hiring into entry level positions. Travenol did not hire blacks at the same rate as whites, comparing applicant flow to hires.

2. Discrimination in initial assignment of those hired. Travenol assigned blacks disproportionately to custodial rather than operative work.

' 3. Discrimination in promotions. Travenol hindered the advancement of women from assembler to material handler, and offered both women and blacks scarcely any opportunity for promotion above the operative level.

4. Discrimination in merit pay. Travenol paid less to blacks and women in positions above the operative level for which the salary is determined subjectively through the merit system.

24 . In Payne I, Payne argued that she had standing to challenge the tenth grade requirement because she lacked a tenth-grade equivalency for 80 of the 90 days preceding her January 29, 1970 EEOC charge. We rejected this argument, reasoning that “the trial court’s definition of the class removed from consideration all applications for employment submitted to Travenol prior to March 3, 1970, so that the situation prior thereto is irrelevant on this appeal.” 565 F.2d at 899.

The opinion in Payne I reveals that the parties did make some reference to the earlier suggested date of January 29, 1970, and a reference back to a period of up to ninety days preceding that date. It must be concluded, therefore, that the issue of the establishment of March 3, 1970, by the district court as the beginning date for applications of members of the class was considered by the Court in Payne I and became as a result of its holding the critical date for the class as it related to the tenth grade requirement. If the members of the class felt at that time that the decision was wrong because the precise date was wrong and that, therefore, they did have members of the class who did not have the tenth grade requirement or its equivalent at a proper date for the class they had the obligation to pursue their position further. They were required to raise that specific issue upon a motion for rehearing to this Court at the end of Payne I. Otherwise, the holding became and still continues to be the law of the case (Conway v. Chemical Leaman Tank Lines, 644 F.2d 1059 (5th Cir. 1981)) on that one issue. The parties cannot now be heard to urge that the tenth grade requirement as it relates to the class can be based upon a ’ commencing date other than the one which was before the Court and upon which the Court clearly relied in making its decision.

This conclusion is buttressed by the fact that the issue of the validity and application of the tenth grade requirement was not raised upon the current appeal and is not before us. We, therefore, have considered the matter as to the proper date for the inception of the class only as it applies to issues other than the invalidated tenth grade requirement. That issue is foreclosed by Payne I.

25 . Burdine holds that an individual seeking to prove discriminatory treatment must establish a prima facie case according to McDonnell-Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). The defendant then has the evidentiary burden of articulating a reasonably specific, legitimate reason for its action. If the defendant meets this burden, the plaintiff then must show that the defendant’s reason is a pretext for discrimination. See McWilliams v. Escambia County School Board, 658 F.2d 326 (5th Cir. 1981).

The Third and the Ninth Circuits have applied Burdine to Title Vll class actions. See Croker v. Boeing Co., 662 F.2d 975 (3d Cir. 1981) (en banc); Piva v. Xerox Corp., 654 F.2d 591 (9th Cir. 1981).

26 . The district court cited no statistics to support this finding of disparate result. We consider later whether this finding is clearly erroneous in light of the statistics submitted by the parties.

27 . The court found that Travenol hired one of three white applicants but only one of six black applicants between April 1971 and August 1974.

28 . The statistics relied on by the court showed the following percentages of whites in the operative work force.

Percentage of the operative work force that was white: Year

All Male Female

100 1964

84 1966

82% 89.8% 1969

76.6% 82.8% 1970

54.1% 70.1% 1973

37.4% 57.5% 1974

29 . Although liability under Title VII cannot rest on Travenol’s pre-Act discrimination, its conduct before 1964 is relevant to the interpretation of post-Act disparate results. See Hazelwood School District, supra, 433 U.S. at 309 n.15, 97 S.Ct. at 2742 n.15.

30 . The factors considered by the third personnel manager were alleged to be “personal cleanliness, general alertness, and comprehension.” No fixed list of questions was used to measure these qualities; the questions varied from applicant to applicant.

31 . Under McDonnell-Douglas the plaintiff may establish a prima facie case of individual disparate treatment by showing: “(i) that he belongs to a racial minority; (ii) that he applied and was qualified for a job for which the employer was seeking applicants; (iii) that, despite his qualification, he was rejected; and (iv) that, after his rejection, the position remained open and the employer continued to seek applicants from persons of complainant’s qualifications.” 411 U.S. at 802, 93 S.Ct. at 1824. By noting that the named plaintiffs satisfied the McDonnell-Douglas criteria, we do not imply that proof of individual instances of discrimination presented to support a case of classwide discrimination must conform to McDonnell-Douglas.

32 . The following chart presents the breakdown of applicants and hires:

Hiring Rates of Applicants Referred by MSES, for Assembler and Material Handler Positions, from November 3,1969Through November 24,1971

White Black White Black Males Males Females Females

Total Applicants Referred 293 138 503 440

Total Referrals Hired 130 29 260 107

— Percent Hired 44.3% 21-0% 51.7% 24.3%

The calculation of standard deviations though not essential, see Rivera v. City of Wichita Falls, 665 F.2d at 545, n.21, is extremely useful. Absent explanation, standard deviations of greater than three generally signal discrimination, see Hazelwood School Dist., 433 U.S. at 311 n.17, 97 S.Ct. at 2743 n.17. Standard deviation analysis here reveals that the discrepancy between black and white hires is probative of discrimination. The observed number of hires is 7.54 standard deviations fewer than the expected number.

The “standard deviation” is a way to calculate the likelihood that chance is responsible for the difference between a predicted result and an actual result. See Castaneda v. Partida, 430 U.S. 482, 496 n.17, 97 S.Ct. 1272, 1281 n.17, 51 L.Ed.2d 498. For example, if we flip a penny one hundred times, we expect that fifty will be heads and fifty tails. If 51 flips out of 100 produce heads, we chalk it up to chance. But long before we get to 99 flips out of 100 producing heads, we suspect the penny. Standard deviation analysis is the mathematical means of expressing how likely it is that the penny is suspect. For hiring discrimination, the standard deviation is calculated by taking the square root of the product of the number of applicants hired times the probability of hiring a minority times the probability of hiring non-minority. In our penny example, the standard deviation is the square root of 100 times .5 (the probability of heads) times .5 (the probability of tails), or 5. If 100 flips yield 40 heads, the observed value falls two standard deviations below the expected value. The probability of this happening by chance is less than five times in 100. Equal Employment Opportunity Comm. v. American Natl. Bank, 652 F.2d 1176, 1192 (4th Cir. 1981). The probability of 100 flips yielding 35 heads — or 3 standard deviations below the expected value — is less than one in 100. Id. Statisticians tend to discard chance as an explanation for a result when deviations from the expected value approach two standard deviations. Id.

33 . The plaintiffs set out the following table:

Applicants With a Tenth-Grade Education, From 1970-74, Compared With Hires Into Operative Positions, 1970-74_

Whites Blacks

Applicants for Operative Jobs With a 10th Grade Education 1,478 2,193

Hires in Operative Jobs, 1970-74 664 548

% Hired 44.9% 25.0%

Calculation of standard deviations reveals that the observed number of black hires is 10.28 standard deviations fewer than the expected number.

34 . The plaintiffs presented the following chart:

Applicants With a Tenth-Grade Education, From 1972-74, Compared With Hires Into Operative Positions, 1972-74_

Whites Blacks

Applicants for Operative Jobs With a 10th Grade Education, 1972-74 636 1,658

Hires in Operative Jobs, 1972-74 166 356

% Hired 26.1% 21.5%

' The disparity, though less than earlier years, is still significant: the number of black hires falls short of the expected value by more than two standard deviations. /

35 . Applications remained current for ninety days in 1974 and later years. Before 1974, applications were considered current for thirty days.

36 . Even if we assume that 25% of the applications were duplicates, Travenol still hired 59.8% of the white applications while hiring only 33.3% of the black applicants between 1970 and 1974.

37 . Travenol also removed duplicate applications in computing these proportions.

38 . The plaintiffs prepared the following chart from Travenol’s EEO-1 reports:

Number of Date of EEO l Job Number of Black Report Categories Employees Employees % Black

November 30, Operatives 1,532 880 57.4%

1977 Service 87 67 77 0%

Workers

November 30, Operatives 1,097 658 60.0%

1978 Service 63 Workers 48 76.2%

November 30, Operatives 1,109 687 61.9%

1979 Service 65 46 70.8%

Workers

39 . The impact of the “hooks" on women was particularly severe because until 1968, Travenol had steered women into the lower paying job of assembler and men into the higher paying job of material handler. The overt sex requirements were dropped in 1968, but the impressions left in the minds of applicants resulted in no males applying for the assembler job and no females applying for the material handler job until 1971. By 1974 there were 630 female and two male assemblers, and 226 male material handlers but only seven female material handlers.

40 . More recent statistics also reveal a concentration of blacks in the lower paid positions in Travenol’s plant. While whites constituted 38.1% of the operative work force in 1979, the following statistics show their proportions in other positions in the plant:

August 1979

Job Level No. % White

Officials & Managers 66 83.5%

Professionals 26 72.2%

Technicians 17 70.8%

Office & Clerical 44 77.2%

Craftsmen 122 68.2%

41 . Without reviewing in detail here all of Payne’s exhibits, we make the following observations on her statistics. Payne states in her brief that “black female Craftsmen were paid an average of almost $600 a year less than white female Craftsmen, and white female Craftsmen were paid an average of $1324 less than white male Craftsmen.” Looking at the exhibits, we are unable to tell how long any of the craftsmen have been working, a fact that is certainly relevant to their salaries. Moreover, we note that there is only one black female craftsman on the exhibit. We recognize that Travenol’s discriminatory promotion policies account for the paucity of black females in merit pay jobs. But the small number of them — one in the category of craftsmen— makes it difficult to prove disparities in pay statistically.

Payne also relies on exhibits for proof of patterns of disparate merit pay. These exhibits show the date of hire, date of hire for the merit pay job, and the sex, race, and salary of each employee in a merit system job category during 1973 and 1974. All but two job categories contain no black females. In one category, inventory clerk, the sole black female is paid more than the average white female; no males are employed as inventory clerks. In the other category, accounting clerks, the sole black female makes twenty two cents an hour less than the average hourly amount made by white females; again, no males held this job. Our review of the chart reveals that only one white female with less seniority than the black female made more than the black female did. The figures for 1974 are more equivocal. In that year, white females received an average raise of thirty five cents an hour; the black female’s raise was only fifteen cents per hour. While this fact is suggestive, it is too slender to support a condemnation of an entire system of setting salaries.

42 . We recognize that February 19, 1976 is used in the decree because the district court found that discrimination had ended by the date it rendered its decision on liability. We hold in II.A, supra, that the district court did not make findings that support a conclusion of discrimination in hiring beyond 1974. We therefore are remanding to permit the district court to make findings that underpin its conclusion that discrimination continued until 1976, or to modify its conclusion. The decree, of course, must reflect any alteration that the district court makes in the ending date for discrimination at Travenol. We use February 19, 1976 in this opinion as a proxy for the date eventually selected by the district court.

43 . The decree provides:

Any class member who is entitled to constructive seniority under the provisions of paragraph 10 as a result of discriminatory delay in her hire, and who has been laid off and not been recalled as of the effective date of this Decree, shall be recalled in accordance with her constructive seniority date and the plant or constructive seniority dates of the other employees laid off. The employees whose seniority is to be compared are the employees in competition for recall with the class members in question.

44 .' The following hypothetical situation illustrates this point: A white female and a black female applied to Travenol in 1970, and the white female was hired immediately but the black female was not hired until 1976. Both were then laid off in 1976. Travenol recalled employees in 1977 in order of seniority, and, because of her greater seniority, the white female was recalled but not the black. In 1980, the black female was awarded a constructive seniority date of 1970. Because the black female should have been recalled in 1977, had she not suffered a discriminatory delay in hire, she would be entitled to compensation for the time she awaited recall.