Defendant American Commercial Marine Service Company, Jeffboat Division (Jeffboat), brought this interlocutory appeal to contest the finding of liability against it in this civil rights suit. Plaintiffs are five African-American former employees of Jeffboat, and they sue in both their individual capacity and as class representatives. The district court found defendants liable for violating various provisions of Title VII and section 1981. We affirm in part, reverse in part and vacate and remand in part.

This opinion marks our second encounter with this litigation, and it may not be our last. Filed in 1977, the case was originally tried before the late Judge Holder, who found for the defendant on all counts. We reversed that decision on appeal, however, holding that "the findings of fact made after trial by the district court [were] insufficient to permit meaningful appellate review." 746 F.2d 365, 367 (7th Cir.1984). We instructed the district court on retrial that it should not feel bound by the law of the case or by any similar analytical constraint. After remand, the case was retried before Judge Parsons.1 The district court issued several thorough opinions on liability, which, on balance, favored the plaintiffs. Among these opinions was an order reaffirming the court's judgment in light of the intervening Supreme Court decisions in Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S. Ct. 2115, 104 L. Ed. 2d 733 (1989) and Patterson v. McLean Credit Union, 491 U.S. 164, 109 S. Ct. 2363, 105 L. Ed. 2d 132 (1989).2 The district court then granted defendant's request to seek interlocutory appeal. Our order allowing the appeal was issued on July 12, 1990.

This case well illustrates the substantial difficulties in maintaining and in defending a statistics-based class-action discrimination suit. We are grateful to the district court for its comprehensive scrutiny of the arguments of both parties. With this more complete treatment of the evidence by the district court, we believe we may take a fresh look at the facts (rather than merely rehash our earlier opinion).

Jeffboat has operated its immense shipbuilding facility on the banks of the Ohio River in southern Indiana since 1938; maritime construction on the property dates at least as far back as 1876. Through the 1950s, Jeffboat operated its facility with an all-white work force. But societal and governmental pressures associated with the civil rights movement brought changes, and by 1970 minorities represented 14.1% of the company's new hires.

As a government contractor, Jeffboat was required by federal law to engage in affirmative-action programs focusing on improving both its hiring and on-site training programs. The first such plan was adopted in 1972 and back-dated so as to run from 1971 to 1973. A second federal affirmative-action plan was later adopted, though not until 1977. In the interim, Jeffboat was obligated by a consent agreement with the Indiana Civil Rights Commission to follow an affirmative-action program in its hiring and promotion practices.

Notwithstanding the adjustment of Jeffboat's hiring practices and its submission to these affirmative-action plans, many minority employees believed that discrimination continued to pervade the facility's operations, affecting certain post-hire practices. Following their termination from Jeffboat, four named plaintiffs filed this suit in 1977: William O. Mozee, Gregory L. Rankin, Frederick Williams and Joe L. Malone. An action by a fifth named plaintiff, Harold Barnes, was filed in 1978 and was consolidated with the earlier case for trial. The court granted the plaintiffs class certification on August 17, 1978.

Plaintiffs have charged the defendant with discrimination and with retaliation in violation of sections 703(a) and 704(a) of Title VII, 42 U.S.C. Secs. 2000e-2(a), 2000e-3(a) (1976),3 as well as with violations of section 1 of the Civil Rights Act of 1866, 42 U.S.C. Sec. 1981 (1976).4 They hoped to prove their Title VII claims of class-wide discrimination by either the disparate treatment or disparate impact methodology. International Bhd. of Teamsters v. United States, 431 U.S. 324, 335 n. 15, 97 S. Ct. 1843, 1854-55 n. 15, 52 L. Ed. 2d 396 (1977) (disparate treatment); Griggs v. Duke Power Co., 401 U.S. 424, 432, 91 S. Ct. 849, 854, 28 L. Ed. 2d 158 (1971) (disparate impact). Only proof of disparate treatment would permit an award under section 1981. General Bldg. Contractors Ass'n v. Pennsylvania, 458 U.S. 375, 391, 102 S. Ct. 3141, 3150, 73 L. Ed. 2d 835 (1982). The plaintiffs chose several areas of post-hire decision-making at Jeffboat for scrutiny: promotions, bid awards, seniority rights, job assignments and various disciplinary determinations. Complaint--Class Action at 36-37. The evidence marshalled by the plaintiffs can be divided into two categories, individual occurrences and class-wide proof.

The district court found that each of the individual plaintiffs experienced discrimination in one or more areas of treatment at the hands of Jeffboat. Defendant does not appeal the factual determinations leading the district court to find it liable with respect to the individual plaintiffs.5 Nevertheless, anecdotal evidence of discrimination may be used to supplement a showing of class-wide disparate treatment, Coates v. Johnson & Johnson, 756 F.2d 524, 532 (7th Cir.1985), and the district court relied on such evidence in this case. Because Jeffboat challenges the sufficiency of the evidence supporting the determination of class-wide discrimination, we will summarize the district court's findings of individual discrimination.

William Mozee charged Jeffboat with discrimination in failing to promote him, failing to satisfy his arbitration award and exercising retaliatory or racially motivated discipline and discharge. Mozee was an overhead crane operator at Jeffboat in June 1975, at which time a higher position became available as a gantry crane operator. The position was to be awarded through the bidding system, which selects among equally qualified employees the candidate with the most seniority. Of the five employees submitting bids for the gantry crane position, three were white and two were African-American. Two of the white employees and the other African-American applicant were eliminated from consideration for reasons irrelevant at this point, leaving Mozee and Earl Cartreete. Despite Mozee's greater seniority compared to Cartreete, the white employee received the promotion.

Mozee went to arbitration, where Jeffboat conceded that the two were equally "qualified" but defended its action on the ground that Mozee's poorer attendance record made him less desirable. The arbitrator found that this justification violated the collective bargaining agreement and ordered Jeffboat to award Mozee the job and retroactive compensation. Jeffboat eventually paid Mozee retroactive compensation but the gantry crane position he sought was eliminated before the company could place him there. The district court agreed with Mozee (and the arbitrator) that awarding the position to Cartreete represented an aberrational promotion practice evidencing an intent to discriminate on the basis of race. The court refused to hold Jeffboat responsible for not satisfying the arbitration award, however, finding no failure to comply.

Mozee was one of many employees who took part in a mass protest against Jeffboat for its alleged discriminatory practices. The August 1975 demonstrations, known as the "Black Days" protests, involved speeches, petition signing and a call for negotiations between the Black Workers' Coalition and Jeffboat's administration. After notifying Jeffboat of his intent to miss work, Mozee participated in both the August 11 and August 21-22 protests. Immediately upon his return after each protest, Mozee was disciplined--first with a five-day suspension and then with the stiffest allowable ten-day suspension. The district court found that under Collins v. Illinois, 830 F.2d 692, 702 (7th Cir.1987), Mozee had made out a prima facie case of retaliation and further found that the disruption caused by his absence did not outweigh the value of carrying out Title VII's policies. It went on to discredit Jeffboat's nondiscriminatory justification as pretextual, noting that white employees with similar avoidable absences had in the past faced far less rigorous penalties. Parker v. Secretary, U.S. Dept. of Housing and Urban Dev., 891 F.2d 316 (D.C.Cir.1989).

Finally, Mozee claimed his discharge in August 1976 was racially motivated. After an extended absence from work, excused for medical and other reasons, Mozee returned on August 2, 1976, but experienced a string of tardies for eleven consecutive days. The record that Jeffboat kept of these delinquencies seemed to reflect disproportionate scrutiny; this and other indicia led the district court to question the treatment accorded Mozee. Comparisons to white employees shown to be similarly situated revealed that whites received much more lenient treatment, even after two or three times as many tardies. The district court determined that Mozee's termination had been racially motivated. (It rejected Mozee's claim that the discharge was retaliatory, believing there to be a failure of proof of causation.)

Frederick Williams, like Mozee, worked for Jeffboat as a rail crane operator in June 1975 and applied for the opening as a gantry crane operator. His bid was rejected because he was deemed unqualified due to a physical defect. Williams' defect, a congenital problem with his left hand, had been in existence at his hiring in 1968 and continued during the various promotions which brought him to his position as a rail crane operator. The defect had not caused any accidents or occasioned any charges of improper operations against him as a rail crane operator. Nevertheless the company cited as justification for disqualifying Williams the possibility that he would not be able to handle the gantry crane's many operating levers during an emergency situation. He was therefore unable to bid for the position, which eventually went to a white worker ranking below him in seniority. Williams' first claim alleged that the disqualification decision was in fact grounded in discrimination.

The district court placed on Williams the burden of proving that Jeffboat's justification for its treatment of him was pretextual; Williams failed. The court held that, while the employer offered no medical evaluation to support its determination, the mere fact that the decision was poorly supported did not necessarily show it to be pretextual. Williams was thus not able to show that Jeffboat's concern centering on the complexity of the crane and the safety of the crew was pretextual.

In Williams' second claim he argued that he was discharged due to his race. On Friday, July 18, 1975, Williams sought an "out-gate" pass to leave work early for personal reasons but was denied. After the mid-shift break in the afternoon, he received a phone call from his wife to return home for an emergency. Jeffboat granted him the pass, but requested verification of the emergency. The following Monday, Williams returned to work without an adequate written excuse. The company discharged him for insubordination in violation of company rule six. Williams won his case before the arbitrator, who awarded him reinstatement but not back pay. He brought the same claim in the district court.

The district court found Williams' termination racially discriminatory. It agreed with Jeffboat that the offense was one sanctionable by termination but went on to find that Jeffboat disciplined many white employees in similar situations merely with suspensions. Because Jeffboat could not defeat this showing of pretext, the district court found for Williams.

Another employee involved in the Black Days protests, Gregory Rankin, was terminated shortly after they took place. In addition to the two days of demonstrations he attended on August 20 and 21, Rankin also missed work on August 22 due to car trouble. Jeffboat was aware before each absence of Rankin's excuse. Nevertheless, the company ordered a discharge hearing on August 27, when Rankin was required to present a written excuse for each of his three absences. Because the company would not accept attending the protest as an excuse for absence, only the car trouble excuse remained viable. Rankin refused to provide a written excuse, and he was discharged. As part of its justification for this discharge, Jeffboat cited two prior instances when Rankin had been discharged but later, on reconsideration, had been reinstated.

Based on its determination of common practice at Jeffboat, the district court found Rankin's discharge to be racially motivated. It found many examples of white employees who had missed work on numerous occasions for "car trouble" without having to face a demand for verification. One such worker experienced fourteen incidents of car trouble without ever needing verification. Even when Jeffboat finally began asking that employee for verification, further incidents continued to go unverified. Another white employee missed three consecutive days without excuse and even phoned in later in the month that he had car trouble--all without need for verification and without being disciplined. The district court also noted that white employees with prior voided discharges usually enjoyed, unlike Rankin, the expungement from the record of those disciplines. Considering all the evidence, the district court held Jeffboat liable for discriminatory and retaliatory discharge.

Joe Malone was employed either as a welder or as a material checker from 1970 until his discharge in 1975. In 1973, Malone began experiencing problems with alcohol. His addiction forced him to take a month-long medical leave in April 1973 and another one in December 1973. It also resulted in a six-day suspension in November 1973 for working while under the influence of intoxicants. Not until May 1975, almost eighteen months after these events, did Malone experience another alcohol-related absence. This time he missed a week at work to obtain detoxification treatment. Jeffboat had notice of his absence on May 13 when he first called in sick, and again on May 15, when his wife informed the company of his in-hospital treatment. On May 16, the hospital called Jeffboat and verified the hospitalization. On his return, Malone was ordered to produce verification, which he did. He was nevertheless fired for extended absence without furnishing a report containing justification.

The district court relied on two factors in concluding that Malone had suffered discrimination on the basis of his race. First, the company changed its proffered justification for firing him several times in the course of considering his grievances. Second, many whites who experienced similar problems with alcohol were treated much more leniently than Malone. Many either had an equal number or more absences but received less serious punishment. In addition, some white employees were allowed back to work on condition of their participating in a professional help program. Hence the court determined that, while Jeffboat's treatment of Malone was understandable in the abstract, it was discriminatory when placed in the context of the company's shifting defenses and its treatment of similarly situated white workers.

Harold Barnes joined Jeffboat on July 31, 1972, eventually working his way up to welder first class. He was discharged on May 14, 1976, following an incident of alleged insubordination on the job. While his shift worked in the yard one afternoon, it began to rain. The union steward complained that work should be halted, but the foreman in his discretion made the decision to continue. Barnes ignored his legitimate options--getting rain gear provided by the company, requesting a medical determination based on a recent bout with pneumonia or asking for an out-gate pass due to illness--and instead simply refused to work in the rain. After a meeting with the personnel office, Barnes was discharged for insubordination.

The union refused to pursue Barnes' grievance. Nevertheless he brought suit against Jeffboat, conceding that he had violated work rule six but presenting numerous accounts of white employees whose similar or more serious conduct was met only with reprimands or suspensions. The district court agreed with Barnes and found Jeffboat liable for discriminatory termination. Apparently the district court believed that Barnes like Mozee and Rankin was a victim of retaliation for his participation in the Black Days.

The plaintiffs presented two types of evidence tending to show class-wide discrimination at Jeffboat. They offered statistical evidence of disparate promotion and discipline practices at the plant, evidence which was probative both of disparate impact and of a pattern or practice of disparate treatment. Cox v. Chicago, 868 F.2d 217, 222 (7th Cir.1989). Plaintiffs also adduced further evidence of a general policy of racial discrimination around the facility, which bolstered their claim that Jeffboat intended to implement a pattern or practice of discrimination. EEOC v. Sears, Roebuck & Co., 839 F.2d 302, 311 (7th Cir.1988).

For their disparate impact claim, plaintiffs isolated as the discriminating employment practice the subjective promotion and discipline system used by Jeffboat. Basing a disparate impact claim on subjective employment practices was expressly accepted by the Supreme Court in Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 990, 108 S. Ct. 2777, 2786, 101 L. Ed. 2d 827 (1988). With respect to promotions, plaintiffs claimed that Jeffboat employed discriminatory subjective promotion practices in appointing both leadmen6 and foremen. The district court analyzed the statistics presented by the plaintiffs, but found sufficient disparity between the races in promotion to make out a disparate impact claim only in promotions to leadmen.7 Using plaintiffs' statistics, the number of African-Americans actually promoted to leadman fell short of the expected number of promotions by 2.13 standard deviations. That figure grew to more than 3 when the statistics included years leading up to the lawsuit. In using plaintiffs' statistics the district court explicitly rejected as unpersuasive defendant's attempt to offer a more refined statistical analysis. We leave recounting of the specifics of this holding to our treatment of Jeffboat's claims on appeal.

Plaintiffs also convinced the district court that defendant's subjective practice of administering discipline and discharge for attendance problems produced a disparate impact on African-American employees. Plaintiffs produced statistics showing that the defendant terminated African-American employees at a much higher rate than would be expected in a random distribution, higher by more than 6 standard deviations. The trial court did not refer to any statistics offered by plaintiff regarding other disciplinary measures. Defendant again attempted through the use of more refined statistics to show that the analysis of involuntary terminations in fact resulted in a disparate impact on African-American employees but for nondiscriminatory reasons. The district court disregarded this proffer both because of the many flaws the court discovered in the compilation of Jeffboat's statistics and because the plaintiffs were able to use these very statistics to bolster their case. This ruling too is challenged on appeal.

Turning to the plaintiffs' pattern or practice case against Jeffboat, the district court considered the statistical proof discussed above as well as the anecdotal evidence presented in the named plaintiffs' individual claims, citing Teamsters, 431 U.S. 324, 97 S. Ct. 1843, 52 L. Ed. 2d 396 (1977). In addition, the court considered nonstatistical proof of class-wide discrimination. For example, Jeffboat employed a practice of stamping each African-American job applicant's file with the initials "DW," a mark which defendant concedes refers to a local African-American civil rights activist named David Wright. While Jeffboat defended this practice as a means of implementing its affirmative action plan, the court found that the mark "added an inference of contempt to the designation of the race of the black employee...." Trial Order at 24-25. The district court also noted that a sign in the production yard, where no African-American worked in the rigging unit until 1973, declared "No Nigger Riggers." Further, the court found that white employees were permitted to request lateral bids while African-American employees were informed that such bids were not authorized.

As additional proof of Jeffboat's intent to pursue a pattern or practice of discrimination, the court pointed to Jeffboat's "dismal" performance in implementing its affirmative action plans. Id. at 130. The court found that Jeffboat failed to inform its supervisors and salaried personnel (those employees responsible for the subjective decisions on promotion and discipline) of these plans and their requirements. The company neglected to establish means by which it could monitor compliance with the plans' objectives. Most surprisingly, the judge found that neither the company's affirmative action officer nor his superior, the vice president in charge of personnel relations, possessed even a basic understanding of how such a plan should be implemented. Id. at 129-30.

Hence the district court found class-wide liability under a disparate treatment theory with respect to Jeffboat's promotion, discipline and discharge practices. The court specifically denied the plaintiffs' claims with respect to discrimination in several other areas of employee relations, including admission into employee training programs, selection of workers for undesirable job assignments and the filling of vacancies through the job bidding system. The court found that the statistics provided by Jeffboat in these areas demonstrated no disparate impact on African-American employees.

Jeffboat has raised a series of challenges both to the district court's factual findings and its conclusions of law. With respect to factfindings, our scope of review is limited, overturning only those findings which appear to be clearly erroneous. EEOC v. Sears, Roebuck & Co., 839 F.2d at 309. On issues of law, of course, our review is de novo.

A. Sufficiency of the Evidence of Class-Wide Discrimination in Promotions to Leadman

Jeffboat takes issue with the district court's determination that plaintiffs have made out a claim of class-wide discrimination in its practices of promotion to the leadman position. We have earlier observed that plaintiffs produced statistical evidence, which supported both their claim of disparate impact and their disparate treatment pattern or practice claim. Preliminarily, we note that defendant does not contest the decision to allow plaintiffs to single out a purely subjective step in the promotion process for disparate impact analysis. Such a use of the disparate impact methodology has been sanctioned by the Supreme Court. "We are also persuaded that disparate impact analysis is in principle no less applicable to subjective employment criteria than to objective or standardized tests." Watson, 487 U.S. at 990, 108 S.Ct. at 2786. Jeffboat's practice of allowing its foremen complete discretion in their choice of whom to promote to permanent and temporary leadmen is properly the subject of either a disparate impact or a disparate treatment inquiry.

Plaintiffs' proof of the statistical disparity between the number of class members promoted to leadman and the number predicted by a random distribution consisted of a comparison of the number of African-American leadmen to the number of African-Americans in the hourly work force. Trial Order at 59. These statistics indicated a standard deviation of more than two. The district judge found that number, when coupled with a standard deviation applicable to temporary promotions to leadman of greater than ten, sufficiently significant to make out a prima facie case of disparate impact. The defendant then took up the burden of refuting those statistics. Allen v. Seidman, 881 F.2d 375, 379 (7th Cir.1989) ("the defendant can always present evidence to show that there was no disparate impact--that it is merely an artifact of the plaintiff's statistical study"); Shidaker v. Carlin, 782 F.2d 746, 750 (7th Cir.1986) ("Defendants may rebut a statistical prima facie showing of disparate impact with statistical evidence of their own....") (citations omitted), vacated sub nom. Tisch v. Shidaker, 481 U.S. 1001, 107 S. Ct. 1621, 95 L. Ed. 2d 195 (1987). The defendant undertook this refutation through statistical refinements indicating an absence of discrimination.8

Specifically, Jeffboat pointed to qualifications which, while not shared by all members of the work force, were, nevertheless, necessary to performance as a leadman. On appeal, Jeffboat argues that the relevant pool must take into account seniority, membership in the craft of the open leadman position and first-class rank within that craft. Appellant's Br. at 26-28. Jeffboat argues that, once these qualifications are considered in selecting the appropriate pool for comparison, disparities in promotion recede or disappear.

In selecting an appropriate pool and performing regression analysis in Title VII cases, the Supreme Court has taught that plaintiffs need not take into account "all measurable variables," but rather must simply include the major factors potentially responsible for any disparity. "Normally, failure to include variables will affect the analysis' probativeness, not its admissibility." Bazemore v. Friday, 478 U.S. 385, 400, 106 S. Ct. 3000, 3009, 92 L. Ed. 2d 315 (1986). We have ourselves understood this principle to require that plaintiffs eliminate " 'the most common non-discriminatory reasons' " for any suggested disparity. Coates, 756 F.2d at 541 (quoting Texas Dept. of Community Affairs v. Burdine, 450 U.S. 248, 254, 101 S. Ct. 1089, 1094, 67 L. Ed. 2d 207 (1981)). The trial judge's selection of the appropriate variables required to determine a suitable pool for comparison purposes is generally a question of fact and is reversible only if clearly erroneous. Sears, Roebuck, 839 F.2d at 309. "Especially where statistical evidence is involved, great deference is due the district court's determination of whether the resultant numbers are sufficiently probative of the ultimate fact in issue." Soria v. Ozinga Brothers, Inc., 704 F.2d 990, 995 n. 6 (7th Cir.1983).

Here the district court rejected the defendant's attempt to limit the pool of employees for comparison purposes, and we cannot say that its decision was clearly erroneous. First, we note that with respect to a promotion system such as this one--where Jeffboat concedes that there were no written criteria for the foremen to follow--the use of hindsight to construct "qualifications" for a position must be viewed with some suspicion. Crawford v. Western Elec. Co., 745 F.2d 1373, 1385 (11th Cir.1984) (expressing doubt over reliability of undocumented, unannounced "framework" for discretionary decisions). The trial judge was correctly dubious of the seductive logic of post-hoc explanation. For example, at trial Jeffboat also claimed that blueprint reading was a requirement most foremen took into account when choosing leadmen. The court disbelieved this assertion, however, when the defendant's own witness testified that such a skill was not a leadman requirement. Trial Order at 62. Where subsequently discovered "qualifications" are not obvious and are far from uniformly applied, we think their use in determining the promotion pool must be carefully restricted.

The district court found that each of the "qualifications" offered by the defendant lacked sufficient probativeness to refute plaintiffs' prima facie case. "[T]he principal [qualification]," argues Jeffboat in its brief on appeal, is "experience as a first-class employee in the particular craft in which the vacancy occurred." Appellant's Br. at 26. Nevertheless Jeffboat concedes that nearly ten percent of those it selected from the at-large hourly work force for promotion to permanent leadman were without this qualification. The district court also noted numerous exceptions to this rule in filling temporary leadman openings, Trial Order at 61-62, positions it found to be filled through the same process as permanent leadman. Id. at 51-52. Because so many employees outside the first class of the relevant craft were selected for promotions or upgrades, the district court found it inappropriate to limit the selection pool as Jeffboat requested. Id. at 61-62. Even as to membership in the same craft as the leadman opening, this "qualification" too had exceptions, especially in temporary promotions. Id.; Appellees' Br. at 10.

Jeffboat also offered employee seniority as a qualification a foreman would consider in selecting a new or temporary leadman. We note that this argument does not appear in Jeffboat's post-trial brief, but rather only in its motion for reconsideration. Defendant's Post-Trial Brief at 37-38; Memorandum in Support of Defendant's Motion for Reconsideration, App. A at 5. Even had defendant properly presented the seniority factor to the district court at trial, we think the qualification would have been rejected. Like blueprint reading, this "requirement" was disavowed by the defendant's own witness. Appellees' Br. at 12. More significantly, African-American employees were senior to the white employees actually promoted in eight of the forty-three promotions to permanent leadman. App. at 278 n. 5. With respect to the further "qualifications" presented to the district court, we do not find the rejection of Jeffboat's attempt to limit the promotion pool to be clearly erroneous. We therefore approve the determination of the district court that plaintiffs presented a prima facie case of class-wide disparate impact in Jeffboat's promotional system.

B. Sufficiency of the Evidence of Class-Wide Discrimination in Matters of Discipline and Discharge

We have already referred to the plaintiffs' prima facie statistical case challenging defendant's discipline system. The centerpiece of this showing was a comparison of the involuntary termination rates of African-American and white hourly workers. This comparison produced a gross disparity, with African-American workers terminated at a rate more than six standard deviations higher than terminations of white employees. Trial Order at 66. Like disparities resulted whether the statistics centered on terminations for absenteeism or for insubordination. Id. The district court pointed to no other statistics offered by the plaintiffs apart from these termination figures.

Defendant responded to plaintiffs' analysis with its own analysis of the discipline process in hopes of producing more pointed statistics that would eliminate the appearance of discrimination. Allen v. Seidman, 881 F.2d at 379. These calculations compared the number of days workers from each racial group were absent or tardy before incurring a first, second or third level offense. The numbers apparently showed that African-American employees actually enjoyed a higher level of tolerance before being subjected to any of the prescribed disciplinary sanctions. Def.Ex. AK-II. The district court rejected this rebuttal for several reasons. First, it questioned the accuracy, and therefore the probativeness, of these statistics due to the many flaws it perceived in their composition. Trial Order at 70-71. It moreover believed that the defendant's statistics were adequately refuted by the plaintiffs, who used the same numbers to show that African-Americans received significantly longer suspensions than whites at comparable offense levels. Id. at 71-72.

Jeffboat concedes, as it must following our earlier opinion in this case, Mozee, 746 F.2d at 372, that plaintiffs may use statistics to prove the disparate impact of an employer's disciplinary system in violation of Title VII. Jeffboat thus set out below to contradict the plaintiffs' comparison with its own statistical analysis. In this appeal as in its post-trial brief, however, Jeffboat argues that, before we even reach its statistical rebuttal of plaintiffs' evidence, we must question whether, standing alone, the plaintiffs' statistics would permit a finding of disparate impact. Jeffboat points out that in Coates v. Johnson & Johnson, 756 F.2d 524, 540 (7th Cir.1985), this court reminded litigants that under Title VII, plaintiffs attempting to make out a prima facie case must take into account the most common nondiscriminatory reasons for the alleged unfair employment practice. See also Allen, 881 F.2d at 378 ("A statistical analysis must cross a threshold of reliability before it can establish even a prima facie case of disparate impact.") (citation omitted). For while it is true here that plaintiffs' statistics show that the termination differential was not the result of mere chance, neither do the statistics point directly to discrimination. The plaintiffs' analysis must take into account the major variables one might expect to cause a statistical disparity. Bazemore, 478 U.S. at 400, 106 S.Ct. at 3008.9 In this case as in Coates, the defendant contends that the prior disciplinary record of the affected members of each class would explain the disparity in terminations. Although the district court did not credit Jeffboat's own statistics because of errors in calculation, those statistics--which classify disciplinary actions not only by offense levels but by extent of wrongdoing prior to discipline--illustrate that Jeffboat's records would allow a more refined inquiry than that provided by the plaintiffs in their prima facie case.

The district court did not require a more searching approach of the plaintiffs: "To dismiss plaintiffs' statistics because of Jeffboat's claim that the statistics fail to account for distinguishing variables would be improper in the absence of evidence showing the particular variables other than race which would distinguish differences in justifiable discipline between white and black employees." We held in Coates that the defendant's offer of termination statistics which took prior disciplinary records into account was probative since it incorporated a nondiscriminatory reason for the disparity shown by the gross figures. Id. at 545. Viewed from another perspective, prior disciplinary records are a proxy for, or a measure of, an employee's quality of conduct over an extended period. This may help establish a comparability of conduct among employees with respect to which one can measure termination experience. In this vein, in the case before us, Jeffboat countered plaintiffs' proof of disparate discipline with statistics that indirectly showed African-Americans having worse disciplinary records over time; statistics that took prior offenses into account thus suggested no disparity in discipline at all between African-American and white employees.10 When an employer's disciplinary system is based in part on a progressive sequence of disciplines, each predicated on both present and earlier offenses, failure to account in the regression analysis for these prior disciplines may detract from the significance of the statistics. Regardless of any inaccuracies and oversights in compiling its statistics, Jeffboat put the plaintiff class on notice that the class had not accounted for at least one of the most well-known nondiscriminatory reasons for disparity--prior disciplinary record, as we have noted, a proxy for, or measure of, prior conduct.11

Plaintiffs did attempt to correct this deficiency, using defendant's own statistics to show that discipline was unevenly meted out even when prior records of discipline were the same. The class showed that punishments imposed on African-American employees for second and third level offenses were significantly more severe on average than those meted out to white employees.12 Nevertheless, the district court relied on this showing only for purposes of calling Jeffboat's statistics into question. In the end the district court apparently placed its essential, if not sole, reliance on the crude statistics offered by the plaintiff class to show disparate impact. In thus relying on an oversimplified showing, the district court erred, and we must therefore vacate and remand for further consideration this finding of discrimination in discipline. A factual dispute continues as to whether, taking past disciplinary records into account, there is a disparate impact on African-Americans with respect to discipline and involuntary termination. The district court is free to consider the question again, accepting additional evidence if necessary.

The plaintiff class argues before us that it should not have to account for prior disciplinary records. It cites Coates for the proposition that a defendant may not offer as a nondiscriminatory explanation for the disparity in plaintiffs' statistics a factor that is itself a subject of discrimination. 756 F.2d at 544. Thus, the class members here argue that they need not consider prior disciplinary records because the earlier discipline was itself meted out in a discriminatory manner. While plaintiffs are correct in citing Coates for the proposition indicated, they ignore the respective burdens on the parties. We held in Coates that it often remains the plaintiffs' burden following the defendant's proffer of a nondiscriminatory factor to show that that factor was itself tainted with discrimination. Id. At least where there are objective measurements of the underlying conduct subjected to punishment (here, the number of days tardy or absent before each successive step of discipline), the defendant has appropriately articulated its reason for believing the explanatory factor is not biased. Id. at 554 (Cudahy, J., concurring). In the case before us the defendant showed that taking records of prior discipline into account eliminated the appearance of disparate impact in the discipline process and asserted without adequate refutation that those records are not the product of discrimination. Plaintiffs can point to no specific finding by the district court, responsive to the defendant's statistical proffer, that the discipline system was inherently biased and therefore ineffective as a nondiscriminatory explanation for the disparate figures. Instead, the class has only the court's broad finding of discrimination in discipline, based so far as we can discern on only the disparity in termination rates. We reject what amounts to an attempt by plaintiffs to bootstrap that proof into both a condemnation of the entire disciplinary system and a justification for disregarding prior disciplinary records as a nondiscriminatory explanation for the gross disparity. On remand the district court may reconsider this matter.

C. Burden of Proof in the Disparate Impact Finding

Jeffboat's third alleged error involves what it believes was an incorrect imposition of the burden of proof in the disparate impact analysis. Title VII allows a "business necessity" defense to a disparate impact showing. This means that an employment practice that results in a disparate effect on a protected group might still survive Title VII if it is sufficiently job-related to constitute a business necessity. Albemarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S. Ct. 2362, 2375, 45 L. Ed. 2d 280 (1975); Regner v. Chicago, 789 F.2d 534, 537 (7th Cir.1986). Prior to 1989, the law in this circuit and elsewhere dictated that, once a plaintiff established a prima facie case of disparate impact, the defendant incurred the burden of proving that there was a business necessity for employing the challenged practice. E.g., Washington v. Electrical Joint Apprenticeship & Training Comm., 845 F.2d 710, 715 (7th Cir.1988); Regner, 789 F.2d at 537. Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S. Ct. 2115, 104 L. Ed. 2d 733 (1989), however, changed that understanding. In that case the Supreme Court announced that "the employer carries the burden of producing evidence of a business justification for his employment practice. The burden of persuasion, however, remains with the disparate-impact plaintiff." Id. at 659, 109 S.Ct. at 2126. The trial on liability here had long since ended when Wards Cove was announced. Nonetheless, in cases like Allen v. Seidman, 881 F.2d at 381, we have applied the new precedent to completed trials and remanded for reconsideration.

Jeffboat maintains that we must likewise vacate and remand this case, for the district court, of course, did not follow the precepts of Wards Cove, but, instead, placed the burden of proving business necessity on the defendant. The defendant offers examples of this incorrect placement of the burden in this case as applicable to Jeffboat's promotion and disciplinary practices. Because of our decision to remand the disciplinary case for further consideration, we forgo treatment of Jeffboat's Wards Cove argument as applied to that area. In the promotions case, defendant complains because the district court "found that Jeffboat had failed to 'establish ... the job relatedness' of the objective qualifications for leadman...." Appellant's Br. at 19 (quoting Trial Order at 63). The objective qualifications referred to in that passage of the court's opinion are the requirements of seniority, membership in the craft unit of the open leadman position and first-class rank within that craft. If these requirements sound familiar, they should. They are the same limitations Jeffboat argues were necessary to limit the pool of workers from which plaintiffs drew their comparison statistics.

Unfortunately for Jeffboat, this similarity is more than mere serendipity; it reveals the fatal flaw in this argument. In its opinion, the district court began with the finding that plaintiffs had singled out a subjective employment practice in challenging the operation of Jeffboat's promotion system, a finding defendant does not contest on appeal. Trial Order at 41-48. The district court based this finding on the fact that Jeffboat's foremen received no formal guidance on qualifications they should require in deciding whom to promote to leadman positions. The defendant did not instruct its foremen with respect to traits, objectively determined, to be considered prerequisite to leadman promotions.

When plaintiffs use the disparate impact method to challenge an objective job requirement--for example, an employment test or a physical requirement--defendant's business necessity defense must focus on the job-relatedness of that requirement. See, e.g., Wards Cove, 490 U.S. at 659, 109 S.Ct. at 2125-26 (citing cases); New York City Transit Authority v. Beazer, 440 U.S. 568, 587 n. 31, 99 S. Ct. 1355, 1366 n. 31, 59 L. Ed. 2d 587 (1979). These cases illustrate that an objective requirement must relate to the ultimate effectiveness of the employee in that position. Plaintiffs then receive an opportunity to show that a less discriminatory alternative is available. Albemarle, 422 U.S. at 425, 95 S.Ct. at 2375. When the lawsuit centers on a subjective practice, however, the nature of the inquiry changes.

In subjective practice cases, plaintiffs argue that it is the very subjectivity of the step in the employment process--taking unquantifiable attributes into account--that causes the discrimination. We allow challenges to such practices because of the unfairness in permitting an employer to perpetuate discriminatory effects by relying for discriminatory results on the individual biases of its managers. Watson, 487 U.S. at 990-91, 108 S.Ct. at 2786-87. It is thus the very subjectivity of the practice that plaintiffs seek to eliminate. In order to sustain a subjective practice through a business necessity defense, therefore, an employer would have to argue that the subjectivity of the step is job-related--in other words, that something about the position requires the selector to make a subjective evaluation of the applicant's abilities. See, e.g., Crawford v. Western Elec. Co., 745 F.2d 1373, 1385-86 (11th Cir.1984) (district court upheld subjective promotion system because "it vests the responsibility for promotion in those who are most knowledgeable in the work of the [employees] and who have a direct responsibility for the proper performance of the installation jobs"; appeals court reversed because company could not support this justification); Zahorik v. Cornell Univ., 729 F.2d 85, 96 (2d Cir.1984) (subjective college tenure decisions deemed job-related because "[a] decentralized decision-making structure founded largely on peer judgment is based on ... the stake of an academic department in such decisions and its superior knowledge of the academic field and the work of the individual candidate"); United States v. Hazelwood School Dist., 534 F.2d 805, 813 (8th Cir.1976) (subjective factors are permissible as long as shown to be job-related), rev'd on other grounds, 433 U.S. 299, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977); see also Bartholet, Application of Title VII to Jobs in High Places, 95 Harv.L.Rev. 947, 974 (1982).

It is true that after Wards Cove Jeffboat should not have the burden of proving that it has a business necessity for employing purely subjective methods in choosing its leadmen. But we need not pursue this inquiry further, for Jeffboat has not even offered an explanation why its selection of leadmen requires subjectivity. Jeffboat has not proffered a reason either at trial (and as reflected in the opinion of the district court) or before us. Instead, as noted, Jeffboat claims that it was error to make it shoulder the burden of proving what requirements its foremen actually applied in choosing a leadman. But, in proffering such requirements, Jeffboat has merely suggested possible limitations on the pool of employees from which the foremen could pick. This is a matter with respect to which the burden is properly on the defendant after the plaintiffs make a prima facie disparate impact case. Allen, 881 F.2d at 379; Shidaker, 782 F.2d at 750. Moreover, Jeffboat's argument that its foremen were in reality using these objective criteria all along actually cuts against any business necessity defense it might offer. If Jeffboat's foremen were simply choosing leadmen who met specific objective standards (despite no formal instruction from the company to do so), then ostensibly leaving the decision entirely to the subjective judgment of those foremen would not appear to be a business necessity. The district court was correct in reaffirming its decision on promotion practice after the Supreme Court's ruling in Wards Cove, since Jeffboat never offered an appropriate business necessity for using a subjective selection procedure.

D. Sufficiency of the Evidence of a Pattern or Practice of Discrimination

The district court found Jeffboat liable under Title VII not only for the disparate impact its employment practices produced but for intentionally engaging in a pattern or practice of discrimination. Proving class-wide liability for intentional discrimination, or disparate treatment, requires a showing by plaintiffs that the employer conducted business through a pattern or practice of discrimination. Teamsters, 431 U.S. at 336, 97 S.Ct. at 1855; Sears, Roebuck, 839 F.2d at 308. More than mere isolated incidents of discrimination, proof in pattern or practice cases requires "that racial discrimination was the company's standard operating procedure--the regular rather than the unusual practice." Teamsters, 431 U.S. at 336, 97 S.Ct. at 1855 (1977) (footnote omitted). Often plaintiff classes make out pattern or practice showings through a combination of strong statistical evidence of disparate impact coupled with anecdotal evidence of the employer's intent to treat the protected class unequally. See, e.g., id. at 337-40, 97 S.Ct. at 1855-57; Sears, Roebuck, 839 F.2d at 310-11. In this case Jeffboat complains that the district court put too much emphasis on the company's flawed performance under its affirmative-action plan. We understand this argument to be part of an insufficiency of evidence contention. Of course, this brings the "clearly erroneous" standard into play. Id. at 309.

The district court grounded its pattern or practice determination in the statistics the plaintiffs produced to support their disparate impact claim while recognizing that anecdotal evidence was equally important to the claim. Trial Order at 122-23; see also Sears, Roebuck, 839 F.2d at 311. Jeffboat's contentions to the contrary notwithstanding, the district court adequately supported its finding with nonstatistical evidence of intentional discrimination. The court did find Jeffboat's noncompliance with its various affirmative-action plans probative of discriminatory intent, as is permitted under Yatvin v. Madison Metropolitan School District, 840 F.2d 412, 415 (7th Cir.1988). Nevertheless, it clearly recognized that such noncompliance is not dispositive of the question: "Evidence of an employer's failure to follow the requirements of an affirmative action plan is also relevant, though not conclusive, in determining intentional racial discrimination." Trial Order at 123 (citations omitted).

This recognition went beyond mere lip service; the court also supported its determination with the anecdotal proof outlined above, namely, discriminatory markings on the employment files of all African-Americans and discriminatory signs posted in working areas. There was more. Strong showings of disparate impact before the jurisdictional time period in this case were relevant to, though not dispositive of, a finding of discrimination. Id. at 52. Lateral bids within the same seniority unit were routinely granted whites but denied blacks. Id. at 132. Beyond all this, the court found support in the individual plaintiffs' experiences at the hands of the company, especially their treatment following the Black Days protests.

With this anecdotal evidence to support its judgment, we cannot say that the district court's finding that Jeffboat was guilty of a pattern or practice of discrimination was clearly erroneous. We have earlier vacated and remanded the aspect of the case involving the statistical analysis of Jeffboat's discipline system. Whatever the result of that remand, the district court must obviously reevaluate as well its finding of class-wide disparate treatment. That finding now rests in part upon what appears to be a strong statistical disparity in terminations. Therefore, the reevaluation of the discipline claim may well have a significant impact on the viability of the class-wide disparate treatment claim.

E. The Effect of Patterson v. McLean Credit Union on Section 1981 Liability

As a supplemental basis of liability, plaintiffs sued under 42 U.S.C. Sec. 1981, which has been held to prohibit private persons from discriminating against certain minorities in private contractual relationships. Runyon v. McCrary, 427 U.S. 160, 96 S. Ct. 2586, 49 L. Ed. 2d 415 (1976); Movement for Opportunity and Equality v. General Motors Corp., 622 F.2d 1235 (7th Cir.1980). The standards for proving section 1981 liability closely parallel those for Title VII, with the notable exception that proof of disparate impact does not support section 1981 liability since intentional discrimination is required. General Bldg. Contractors, 458 U.S. at 391, 102 S.Ct. at 3150. Nevertheless, the district court found disparate treatment, and therefore section 1981 liability, in Jeffboat's treatment of the individual plaintiffs, as well as in the company's class-wide handling of promotions and discipline.

After trial in the case before us, the Supreme Court decided Patterson v. McLean Credit Union, 491 U.S. 164, 109 S. Ct. 2363, 105 L. Ed. 2d 132 (1989). In Patterson the Court clarified the scope of section 1981 by limiting its protection to "conduct at the initial formation of the contract and conduct which impairs the right to enforce contract obligations through legal process." Id. at 179-80, 109 S.Ct. at 2374. That decision and its progeny obviously impact heavily on plaintiffs' section 1981 claims. Following post-trial supplemental briefing on Patterson's effect, the district court reaffirmed each of its liability findings. Patterson Order at 27.

This circuit has interpreted Patterson to apply retroactively and to exclude from section 1981 coverage actions involving wrongful terminations or other disciplinary procedures. McKnight v. General Motors Corp., 908 F.2d 104, 108-09 (7th Cir.1990), cert. denied, --- U.S. ----, 111 S. Ct. 1306, 113 L. Ed. 2d 241 (1991). In addition, Patterson circumscribed section 1981 actions for failure to promote. Consonant with its understanding that the statute covers only the making and enforcement of contracts, the Court held that "[o]nly where the promotion rises to the level of an opportunity for a new and distinct relation between the employee and the employer is such a claim actionable under Sec. 1981. Cf. Hishon v. King & Spalding, 467 U.S. 69 [104 S. Ct. 2229, 81 L. Ed. 2d 59] (1984) (refusal of law firm to accept associate into partnership) (Title VII)." Id. 491 U.S. at 185-86, 109 S. Ct. at 2377. In addition, this circuit has also interpreted Patterson to restrict retaliation claims, allowing such claims only when they involve retaliation for attempting to enforce contractual rights. McKnight, 908 F.2d at 111-12.

The effect of these developments on the case before us is substantial although, in some respects, not entirely clear. Thus, regardless of the outcome of our remand on the plaintiff class' pattern or practice claim for discriminatory discipline, Patterson and McKnight preclude recovery for these claimed injuries under section 1981. Similarly, the individual plaintiffs' claims for wrongful suspension or termination no longer involve section 1981 liability. Plaintiffs' individual retaliation claims and their individual and class-wide promotion claims require closer scrutiny.

With respect to the successful retaliation claims of Mozee, Rankin and Barnes, McKnight provides guidance beyond that available to the district court when it issued its post-Patterson order.

[Section 1981] punishes "efforts to impede access to the courts or obstruct nonjudicial methods of adjudicating disputes about the force of binding obligations, as well as discrimination by private entities, such as labor unions, in enforcing the terms of a contract." ... If, therefore, [the company] had, because of [the plaintiff's] race, obstructed his efforts to enforce his contractual entitlements, for example by retaliating against him for bringing suit to enforce his contract of employment with [the company], the company might be guilty of violating section 1981.

... The [plaintiff's] complaints were efforts to enforce his rights under antidiscrimination laws, which is a different matter because "the right to enforce contracts does not ... extend beyond conduct by an employer which impairs an employee's ability to enforce through legal process his or her established contract rights."

McKnight, 908 F.2d at 111-12 (quoting Patterson, citations omitted).

To frame the issue concisely, then, we must inquire whether the plaintiffs experienced retaliation with respect to an effort to enforce contract rights (as opposed to statutory anti-discrimination rights) and whether any such retaliation occurred because of the plaintiffs' race. Plaintiffs thus face a daunting task if they wish to preserve their retaliation awards under section 1981. The instances of retaliation found by the district judge each involved predominantly the plaintiffs' participation in the Black Days, a series of protests over the unequal treatment of African-Americans at the plant. Equal treatment for African-Americans at Jeffboat was not merely a statutory right; it was guaranteed by their collective bargaining agreement, which promised no discrimination by the company on the basis of color. Patterson Order at 12. Thus the first step in the McKnight test is met. Plaintiffs claim retaliation for attempting to enforce contractual rights. Were we required to reach it, we would likely be unequipped to resolve the second question, whether Jeffboat handled these challenges to its disregard of workers' contractual rights any differently because plaintiffs were African-American. The district court made no specific findings to this effect.13

There are, however, other factors which bear on the need for a remand. Although the Supreme Court in Patterson did not face the issue of retaliation under section 1981, it would presumably, based on Patterson, dismiss the claims before us because Jeffboat did not retaliate against any attempt by the plaintiffs to "enforce" their contracts. Writing for five justices, Justice Kennedy described section 1981's prohibitions:

The second of these guarantees, "the same right ... to ... enforce contracts ... as is enjoyed by white citizens," embraces protection of a legal process, and of a right of access to legal process, that will address and resolve contract-law claims without regard to race.... It also covers wholly private efforts to impede access to the courts or obstruct nonjudicial methods of adjudicating disputes about the force of binding obligations ... in enforcing the terms of a contract.

Patterson, 491 U.S. at 177, 109 S.Ct. at 2373. This is a very limiting definition of "enforcement" of contracts and must be construed to apply to retaliation as well as other types of claims. Justice Kennedy's references to "legal process," "access to courts" and "nonjudicial methods of adjudicat[ion]" suggest a considerably narrower interpretation of section 1981's term "enforce" than might previously have been understood.14 In the three instances in which the district court found retaliation in this case, it apparently did so due to the proximity of the discipline to each plaintiff's participation in the Black Days. Trial Order at 138-40. These plaintiffs therefore did not appear to be seeking access to legal process, nor were they pursuing some nonlegal, adjudicatory solution to their complaints. Thus, while dictum in that case, Patterson suggests that the Supreme Court would reject a section 1981 award based on the type of retaliation presented in the case before us. Whatever might be said about the employees' attempts to receive the equal treatment promised them in the collective bargaining agreement, the Black Days, for perhaps obscure reasons, were not, strictly speaking, efforts to "enforce" plaintiffs' contract rights under section 1981 as characterized by the sweeping dictum of Patterson. Thus, unless the retaliation awards under 1981 can be attributed to something other than participation in the Black Days, they must be reversed. It is conceivable that retaliation resulted from something other than, or in addition to, the Black Days participation, but we are unable to identify any finding which would point even remotely to other bases of retaliation. We are therefore driven to the conclusion that the section 1981 retaliation claims of Mozee, Rankin and Barnes, as well as the plaintiff class, must be reversed.

Patterson's final impact on this case involves the plaintiffs' individual and class actions for failure to promote. Since Patterson, this court has on at least three occasions addressed the issue whether a new position represents "a new and distinct relation between the employee and the employer." In Malhotra v. Cotter & Co., 885 F.2d 1305 (7th Cir.1989), we dealt with an accountant whose employer refused him promotions to positions such as finance manager and accounts receivable manager. In that case we suggested an interpretation of Patterson that would bar actions when the plaintiff sought a "routine advancement that only existing employees qualify for." Id. at 1305. This rule might leave section 1981 as a remedy when, for example, the position sought was instead one for which a stranger to the firm could apply.

[The interpretation avoids] the anomaly created by a rule that a stranger to the firm could sue under section 1981 if his application for a position was turned down on racial grounds but a person already employed by the firm could not sue even though his application for the identical position was turned down on the identical grounds.

Id. Malhotra was remanded for further fact-finding on this point.15

Later, in McKnight, we encountered an employee who had been laid off, allegedly due to race, and then retained but in a different division after he filed a discrimination action. While not dealing with a promotion in that case, we again suggested the functional analysis outlined in Malhotra. 908 F.2d at 109-10. Most recently, in Bailey v. Northern Indiana Public Service Co., 910 F.2d 406, 410 (7th Cir.1990), we were able to avoid the question because the plaintiff failed even to present a prima facie case of discrimination. Courts outside this circuit have suggested a fact-specific inquiry into the overall effect the promotion has on individual components of the plaintiff's employment.16

The case before us involves promotions to two positions, gantry crane operator and leadman. The district court found each position represented a new and distinct relation between employee and employer. Patterson Order at 5-8. We review the factual components of this decision for clear error. Nevertheless, many of the reasons the court offered to support its finding involved the manner in which an employee could attain or keep either position. The district court discussed the company's unfettered discretion in promoting or demoting such employees. These reasons are inapposite to the new and distinct relation inquiry, which focuses on the responsibilities of the new position vis a vis the old one, not the process by which the employee is promoted. As for relevant matters considered by the district court, it found the gantry crane position distinct from other crane positions--aside from the pay increase--only in that directions for that machine's movement came from the plant superintendent rather than the foreman and that its performance affected the work of more plant workers. Id. at 7-8. Relevant considerations for the leadman position included only that the leadman possessed some supervisory authority and that the position was a proving ground for foreman selection. Id. at 5-7; Trial Order at 13.

The promotions denied in this case would not have created new and distinct relations between the employees and Jeffboat. The district court did not find that an outsider would be considered by the company for either position. Trial Order at 50. Gantry crane operators were chosen by the bidding process among current employees with on-the-job rail crane experience, Patterson Order at 7, and leadmen were chosen from within the work force as well. Trial Order at 11-12. Nor did the positions represent so significant a change in the employee's duties and compensation that they qualified as new and distinct relations. Neither move would have placed the employee in a salaried position, substantially altered the role he played in operations or placed him outside the collective bargaining unit. Even the leadman position, with its attendant supervisory function, could only fairly be said to constitute a change in the employee's relation to other members of his bargaining unit, not in relation to his employer. Cf. Harrison v. Associates Corp. of N. Am., 917 F.2d 195, 198 (5th Cir.1990) (promotion from C.R.T. operator to lead C.R.T. operator inadequate to meet Patterson's "new and distinct relation" test). Plaintiffs' recovery under section 1981 for Jeffboat's failure to promote must therefore be reversed.

Jeffboat's last allegation of error is that the district court wrongly refused to redefine the class. The company requested that the class be redefined to reflect the absence of any evidence of discrimination or disparate impact after 1978. We need not reach this point at this time. The question of class membership is better left until the district court concludes the damages portion of the trial. At that time its determinations of those employees able to recover under the proof adduced at trial will be appropriate for consideration on appeal.

The claims against Jeffboat by the individual plaintiffs for violation of Title VII are AFFIRMED; those individual claims under section 1981 are REVERSED. The class claims against Jeffboat for Title VII violations in its promotion practice--both through disparate impact and pattern or practice liability--are AFFIRMED; those claims regarding discipline are VACATED and REMANDED with instructions. The class claims under section 1981 are REVERSED.