JOHNSON, Circuit Judge:
This is an appeal from a judgment rendered in favor of the employer in a class action brought under Title VII, 42 U.S.C. § 2000e et seq. Plaintiffs-appellants challenge the promotion practices for field examiners and field attorneys in the regional offices of the National Labor Relations Board (NLRB). After a lengthy trial, the district court entered judgment against the plaintiffs on both the class and individual claims. The district court also awarded costs to the defendants. We affirm the judgment against the plaintiffs on both the class and individual claims but remand on the award of costs to the defendants.
THE CLASS CLAIM
I. BACKGROUND
A. The Parties
Donald R. Lewis, the named plaintiff, has been employed since 1971 as a field examiner in the NLRB’s regional office in Houston, Texas. The defendant is the NLRB. The class which the named plaintiff represents consists of a class of:
Past, present and future black field examiners and attorneys throughout all of the NLRB Regions within the United States who could have brought a timely administrative claim on the date the Plaintiff brought his claim.
Findings of Fact at I. The court defined the class temporally to include only those persons who could have brought an administrative claim within thirty days of the date Lewis brought his claim, which was on September 2, 1975. The district court also restricted the class to persons employed as of December 31, 1980.
B. The Organization of the NLRB
The NLRB is a federal agency established in 1935, which is charged with the duty of enforcing the National Labor Relations Act. The NLRB is composed of a five member Board, a General Counsel, and thirty-three regional offices. At the time this litigation commenced, these thirty-three regional offices employed approximately 1,800 persons.
The office of the General Counsel is divided into four divisions. The Division of Operations Management manages the regional field offices for the General Counsel. The Operations staff is divided into six districts, each of which is headed by an Assistant General Counsel. Each district oversees five to six of the regional field offices.
Each regional office is headed by a regional director, who is assisted by an assistant to the regional director, a regional attorney, and supervisors. These officers supervise a professional staff consisting of field attorneys and field examiners. Examiners investigate allegations of unfair labor practices, investigate cases involving questions of union representation, and conduct union representation elections. Field attorneys have the same duties as field examiners but also try cases in administrative and federal district courts.
C. Initial Hiring and Promotion Tracks of NLRB Professional Employees
1. Field Examiners
The normal entry level for field examiners during the period covered by this lawsuit was at a classification known as GS-7, though some field examiners enter at GS-5. In order to move up the career ladder to the next level, a field examiner must serve one year in grade and receive satisfactory performance evaluations. Field examiners can also be hired at a higher level, GS-9, if the applicant has completed a masters degree in labor relations, industrial relations, or labor economics or if the applicant has special experience. From the entry level, a [1269]*1269field examiner’s career ladder progresses up through GS-12, which is the “journeyman” level. Once a field examiner is hired, the field examiner receives promotions without applying for them and without competition until he or she reaches the GS-12 journeyman level. Regional directors make decisions as to promotions up to GS-12.
Field examiners beyond the journeyman level may be supervisors of other NLRB professional employees, or they may occupy nonsupervisory “expert” positions. Promotions beyond the GS-12 journeyman level include competitive positions such as Compliance Officer, Group Supervisor or District Specialist. A field examiner must ordinarily serve sixteen months at the GS-12 journeyman level before becoming eligible for a promotion beyond the journeyman level to GS-13. A GS-13 examiner is considered to be at an “expert” level position and is expected to handle the most complex cases with a minimum of supervision. In order for a field examiner to be eligible for a higher level, GS-13 position, he or she must be recommended qualified for the position. Unlike promotions up to the GS-12 journeyman level in which promotion decisions are made almost exclusively within the region, recommendations for promotions to competitive positions beyond GS-12 depend on an appraisal process which originates in the regional office and which is reviewed in Washington.
2. Field Attorneys
The normal entry level for field attorneys is GS-9. Field attorneys are also evaluated on an annual basis, and assuming satisfactory performance, can be promoted at one-year intervals up to GS-13, the journeyman level for field attorneys. Regional directors have the authority to promote field attorneys up to the GS-13, journeyman level. The GS-14 field attorney level is expected to handle complex cases with a minimum of supervision. The GS-14 field attorney and above may supervise other NLRB professional employees or may serve as an expert nonsupervisory attorney. The normal waiting period for promotion to GS-14 attorney is twenty-four months. Promotion beyond GS-13 again depends on availability of positions and ratings received through the appraisal process. Like promotions above the journeyman level for field examiners, recommendations for promotion to competitive positions beyond GS-13 field attorney are obtained through an appraisal process, originating in the regional office and reviewed in Washington.
D. The Appraisal Process
1. Appraisals to Journeyman Level
Since 1967, professional employees below the journeyman level have been evaluated on a standardized form which consists of two parts. Part I lists twelve factors on which to evaluate field examiners and attorneys in grades GS-5 through GS-11, and an additional eight factors on which to evaluate attorneys and examiners in GS-12 and above. Next to each factor are several descriptive statements from which the employee’s supervisor may choose to describe the employee’s performance on that factor. The supervisor may also choose not to select any of those statements. The statements are not arranged from least favorable to most favorable (or vice versa), since the developers of the form believed that a varied pattern of options on the form would encourage supervisors to study the form and make more deliberate decisions on which of the descriptive statements to select. Part II of the form is headed “Narrative Comments.” The form states that the supervisor is to comment specifically on the performance of the employee, both strengths and weaknesses, and is to give concrete examples. Supervisors have received training and examples to follow in the appraisal process at least since the mid-1970’s.
Once the supervisor has prepared the form, the employee is allowed to comment on the appraisal before it is sent to the regional director, the assistant regional director, and the regional attorney. These officials determine whether the employee [1270]*1270will be promoted to the next level. The employee has the right, under the collective bargaining agreement between the employees’ union and the NLRB, to meet with the regional director to discuss the appraisal. The employee can also submit written comments on the appraisal. The employee can also challenge the appraisal through the grievance/arbitration procedures provided in the collective bargaining agreement.
2. Appraisals of Journeyman-Level Professional Employees
Evaluations for promotions beyond the journeyman level to competitive positions at either a supervisory or nonsupervisory expert level begin at the regional office and are reviewed at NLRB headquarters in Washington. Evaluations for employees are initially prepared by their supervisors. The employee is allowed to discuss the appraisal with his or her supervisor, meet with the regional director, and attach comments to the appraisal before it is sent to NLRB headquarters. The evaluations of these journeyman-and-above employees must specifically include an appraisal of potential for promotion to the supervisory positions. The regional director recommends whether the employee is “well qualified” or “not ready” for promotion to supervisory positions. The collective bargaining agreement also requires a determination of whether the employee should be promoted to nonsupervisory positions above the journeyman level. This appraisal by the regional director is reviewed in Washington by the Divisions of Operations Management (“Operations”). If the employee disagrees with the recommendation from the regional office, he or she may file comments with Operations.
A panel of three Operations staff members reviews the region’s initial recommendations. The panel consists of the Assistant General Counsel for the district which includes the region, a Deputy Assistant General Counsel, and a third officer, the Executive Assistant. The panel reviews the appraisal and the panel makes the final recommendation as to whether the employee should be considered for promotion to competitive positions as a supervisory or a nonsupervisory professional employee. Employees, as in promotions for lower-level positions, have grievance and arbitration rights.1
E. Promotions Above the Journeyman Level
Even if an employee is rated ready for promotion to a supervisory or expert position, the promotion does not automatically follow since actual promotion depends on the availability of positions.
For actual promotion to the expert non-supervisory position, an employee might be blocked by the “50/20 rule.” In 1969 and 1970, the Civil Service Commission issued a report that there was only enough expert level work to support a percentage of 50 percent of the field examiners at the GS-13 expert level and a percentage of 20 percent of the field attorneys at the GS-14 expert level. The NLRB then decided that no more than 50 percent of the nonsupervisory field examiners and 20 percent of the field attorneys in any single regional office could be placed at the expert level. (The percentage for expert level attorneys was raised to 30 percent in 1977.) Thus, since the inception of the 50/20 rule, employees rated ready for promotion to the nonsupervisory expert level positions have been forced to wait varying lengths of time. The NLRB has mitigated the harshness of the 50/20 rule through transfers and “quality within grade” increases for blocked employees.2
[1271]*1271Similarly, a rating of “well qualified” for a supervisory position does not lead automatically to such a promotion. The collective bargaining agreements provide that first consideration for a vacancy goes to persons rated well qualified in the office where the vacancy occurs, as well as to persons not in the office but who are rated well qualified and who have asked to be considered for supervisory vacancies of the particular office in which the vacancy occurs. If consideration of these employees leads to a minimum number of candidates, the selection may be made without soliciting other applicants. If not, the position is advertised throughout the NLRB. The General Counsel of the NLRB makes the final selection.
II. THE MERITS OP THE CLASS CLAIM
Plaintiffs contend that the statistical evidence presented at trial, together with anecdotal evidence, established a prima facie case of discrimination which defendants failed to rebut. They contend that the district court erred in placing upon them the burden of explaining the speculative reasons suggested by the NLRB for statistical disparities in the promotion of black professionals at the NLRB. After setting forth the basic principles of Title VII litigation in this context, we will examine the district court’s analysis of the statistical and anecdotal evidence presented at trial.
A. Principles of Title VII Litigation in this Context
This Court’s recent precedent establishes that a challenge to an allegedly discretionary promotion procedure fits within the “disparate treatment” analysis, rather than the “disparate impact” analysis under Title VII. Carroll v. Sears, Roebuck & Co., 708 F.2d 183, 188-89 (5th Cir.1983); Pouncy v. Prudential Insurance Co. of America, 668 F.2d 795, 800 (5th Cir.1982).3 Plaintiffs in the instant case challenge the NLRB’s discretionary appraisal and promotion procedure. Accordingly, the district court properly considered the NLRB promotion procedures in light of the disparate treatment analysis.
Disparate treatment analysis, unlike disparate impact analysis, requires a showing that the employer acted with a discriminatory intent. In a class action challenging an employer’s entire promotion procedures, plaintiffs must “establish by a preponderance of the evidence that racial discrimination was the [employer’s] standard operating procedure — the regular rather than the usual practice.” See International Brotherhood of Teamsters v. United States, 431 U.S. 324, 336, 97 S.Ct. 1843, 1855, 52 L.Ed.2d 396 (1977). Statistical evidence is highly relevant in a disparate treatment ease challenging an employer’s promotion practices. In a proper case, a court may infer racial discrimination if gross statistical disparities are shown in analyzing relevant statistics. Carroll, 708 F.2d at 190; Pouncy, 668 F.2d at 802. This Court has noted, “While gross statistical disparities may alone establish a prima facie case of employment discrimination in a proper case, the Supreme Court has cautioned ‘that statistics are not irrefutable; they come in infinite variety and, like any other kind of evidence, they may be rebutted. In short, their usefulness depends on all of the surrounding facts and circumstances.’ ” Wilkins v. University of Houston, 654 F.2d 388, 395 (5th Cir.1981), vacated and remanded, 459 U.S. 809, 103 S.Ct. 34, 74 L.Ed.2d 47 (1982), aff'd in relevant [1272]*1272part on remand, 695 F.2d 134 (5th Cir. 1983) (quoting Teamsters, 431 U.S. at 340, 97 S.Ct. at 1856-7). In probing discriminatory intent, a trial court may examine the history of the employer’s practices,4 anecdotal evidence of class members, and the degree of opportunity to treat employees unfairly in the appraisal process. Carroll, 708 F.2d at 190.
In evaluating the district court’s conclusion that the NLRB’s promotional practices were not the product of discriminatory intent, this Court’s “appellate review centers on the fact finder’s ultimate conclusion as to the employer’s motivation.” Parson v. Kaiser Aluminum & Chemical Corp., 727 F.2d 473, 476 (5th Cir.), cert. denied, - U.S.-, 104 S.Ct. 3516, 82 L.Ed.2d 824 (1984). The district court’s finding on “the ultimate issue, that of discrimination vel non, ... is subject to the ‘clearly erroneous’ standard of Fed.R.Civ.P. 52(a).” Williams v. Southwestern Bell Telephone Co., 718 F.2d 715, 718 (5th Cir.1983) (per curiam). See also Parson, 727 F.2d at 475-76.5
Plaintiffs’ statistical evidence may be broken into two sets: those statistics related to the rates of promotion for professional employees at various GS levels and those statistics related to promotions to supervisory positions.
B. Statistics Regarding Career Ladder Promotions for Professional Employees
Both plaintiffs and the NLRB presented evidence regarding promotion waiting times. These statistics compared black and nonblack professional employees at various GS levels and observed the percentage promoted and the amount of time taken in obtaining a promotion to the next GS level.
Plaintiffs’ expert, Dr. Bruce Levin, introduced a “survival” analysis. Dr. Levin’s analysis aimed at comparing the promotion waiting times for blacks and nonblacks. He did this by taking employees at each of the GS levels for field examiners and field attorneys and observed how long employees remained at that GS level.6 In doing so, Dr. Levin testified that he found a consistent pattern of black field examiners and field attorneys taking more time, to promotion than their nonblack counterparts. The results were statistically significant in three levels (GS-5 and 11 for field examiners and GS-12 for field attorneys) of the eleven examined by Dr. Levin for the 1972-81 period. For promotions for the post-1976 period, results were statistically significant at GS-5 and 11 for field examiners and GS-9 for field attorneys.7
The NLRB introduced statistical evidence to counter the plaintiffs’ statistical [1273]*1273showing. It introduced the testimony of Dr. James Beckett, II. Dr. Beckett performed two different types of analysis, the second of which is more relevant here.8 In his second analysis, Dr. Beckett imitated Dr. Levin’s methodology. Dr. Beckett’s analysis, however, differed in that he removed persons from the data that were not members of the class. Specifically, he removed persons from the data who had entered the NLRB through training programs and student internships.9 Dr. Beckett’s repeated survival analysis yielded statistically significant disparities between blacks and nonblacks in waiting time for promotion in only the GS-11 level for field examiners (for both periods 1972-80 and 1975-80) and in the GS-12 level for field attorneys (for the 1975-80 period only). The formerly statistically-significant results in the GS-5 level for field examiners and GS-9 level for field attorneys faded.10 Dr. Beckett testified that while approximately 85 percent of blacks and nonblacks progressed at the same rate, the differences between the promotion rates for blacks and nonblacks were due to a group of “slow-promoters.” A greater proportion of these slow promoters were black, rather than nonblack. Taking the two levels of GS-11 field examiner and GS-12 field attorney (the levels which showed statistical significance in Dr. Beckett’s repeated analysis) for the 1975-80 period, Dr. Beckett removed the four slowest blacks in both of the two levels. The statistical significance again faded. For each of these eight employees (four black field examiners and four black field attorneys), the NLRB proffered legitimate, nondiscriminatory reasons for their delay in promotion.
The district court accepted this analysis, as well as the NLRB’s explanations for the slow promoters, and concluded that this statistical evidence did not establish discrimination. See Findings of Facts Nos. 64, 66, 68, 69; Conclusions of Law Nos. B.11, B.12. From the contrasting evidence presented by plaintiffs’ and the NLRB’s experts, the district court specifically noted in Conclusions of Law Nos. B.11, B.12:
11. Plaintiff’s statistics regarding waiting time for promotions show no difference between Blacks and Whites as to “fast” promoters (i.e., those especially competent individuals promoted as quickly as possible). A discrepancy did appear as to promotion time for a few “slow” promoters (i.e., those individuals not ready for promotion in the usual length of time).
12. To the extent the statistical evidence may have supported an inference of discriminatory treatment in promo[1274]*1274tions in a few cases, Defendant was able to articulate reasons for a handful of specific instances, and Plaintiff was not able to show that such reasons were pretextual.
It cannot be said that the district court clearly erred in doing so. The NLRB’s effort was aimed at showing that while there were some disparities due to the slow promoters, legitimate reasons, other than nonlegitimate discriminatory ones, were behind those disparities.11
C. Statistics Regarding Promotions to Supervisory Positions
Plaintiffs contend that the district court erred in its evaluation of statistics. According to plaintiffs, these statistics indicate a pattern of discrimination in the promotion of NLRB professional employees to supervisory positions. Plaintiffs rely on a mathematical formula suggested in Hazel-wood School District v. United States, 433 U.S. 299, 97 S.Ct. 2736, 53 L.Ed.2d 768 (1977). Plaintiffs contend that application of the formula shows a statistically significant disparity between blacks and non-blacks in promotions to supervisory positions. Plaintiffs argue that while blacks made up 10.5 percent of the NLRB’s professional work force in 1979, only 4.1 percent of the NLRB's supervisors are black.12 Crucial to the plaintiffs’ contention is that the 10.5 percent figure for black professional employees in 1979 represents the relevant figure for comparison. Plaintiffs note that since the NLRB promotes from within the organization, the relevant figure for comparison is the employer’s (i.e., the NLRB’s) work force, rather than the general population. See Payne v. Travenol Laboratories, Inc., 673 F.2d 798, 826 (5th Cir.), cert. denied, 459 U.S. 1038, 103 S.Ct. 451, [1275]*127574 L.Ed.2d 605 (1982). Since plaintiffs contend professional employees at the NLRB possess “the minimum qualifications for promotion to higher level supervisory and managerial positions at the NLRB,” plaintiffs conclude that the racial composition of the NLRB’s professional work force (regardless of GS level) is the relevant figure for comparison. Plaintiffs-Appellants’ Brief at 46.
The district court concluded that the plaintiffs’ evidence was “not meaningful ... in that it does not provide data regarding the number of qualified, Blacks in the pool from which selections were made for supervisory and other managerial positions.” Conclusions of Law No. B.9 (emphasis added).
This Court has noted “[statistical evidence ... must be meaningful.” Pouncy, 668 F.2d at 802. In establishing an inference of discrimination from statistical evidence the “required comparison [is] to a qualified pool of employees presumptively eligible for promotion.” Id. at 803 (emphasis added). Plaintiffs’ presentation is not meaningful in at least two respects. First, the entire professional work force of field attorneys and examiners is not the relevant point of comparison since not all of these employees are “presumptively eligible” for promotion to supervisory levels. Instead, employees are not even rated for consideration until they reach the GS-12 field examiner level or the GS-13 field attorney level.13 By asserting that the minimum qualification for promotion to higher level supervisory positions is the holding of any professional position, the plaintiffs bring into their comparison lower level employees who are not to be considered for promotion.14 Cf. Pegues v. Mississippi State Employment Service, 699 F.2d 760, 770-71 (5th Cir.), cert. denied, - U.S. -, 104 S.Ct. 482, 78 L.Ed.2d 679 (1983) (plaintiff’s statistical evidence flawed in failing to take relevant experience into consideration). Second, plaintiffs’ selection of the 10.5 percent figure for 1979 is questionable as a basis for comparing promotions to supervisory positions for the period of this litigation (1975-1980) since the number of professional black employees had risen in the immediately preceding years.15 Thus, the district court did not err in finding the plaintiffs’ statistical evidence for supervisory positions unpersuasive.16
[1276]*1276Other facts support the district court’s conclusion that the statistics failed to establish discrimination in promotion to supervisory levels. There were no statistically significant differences in the promotion rates or waiting times for the GS-12 field examiner or the GS-13 field attorney level (the minimum levels for promotion to supervisory positions) for the post-1976 period.17 Further, unlike the situation presented to this Court in the Carroll case, the number of black supervisors increased steadily during the period covered by this suit.18 See Carroll, 708 F.2d at 193 n. 10. The district court’s analysis and conclusions regarding the statistical evidence were not clearly erroneous in failing to find an inference of disparate treatment.
D. District Court’s Analysis of the Appraisal System
Plaintiffs also argue that the “uncontradicted expert testimony presented by plaintiffs by its industrial psychologist demonstrated that the rating systems used were highly subjective and capable of discrimination.” Plaintiffs-Appellants’ Brief at 47. In Conclusion of Law No. B.14, the district court correctly noted that this Court has found appraisal systems which allow for subjective evaluations by white supervisors to be evidence of discrimination in disparate treatment cases. See Payne v. Travenol Laboratories, Inc., 673 F.2d at 826; Rowe v. General Motors Corp., 457 F.2d 348, 359 (5th Cir.1972). The district court, however, also noted several safeguards present in the NLRB appraisal process, which distinguished the appraisal system in the instant case from those criticized in earlier cases. These included review by supervisors other than the employee’s immediate supervisor, the opportunity for employee review and comment of the appraisal, discussion of the appraisals between the employee and NLRB supervisors, and the use of grievance mechanisms by employees to challenge appraisals which employees thought unfair. The district court also noted that it was the NLRB’s policy and practice to train supervisors in evaluating employees and that written guidelines for evaluating employees had been issued since the mid-1970s. Record Vol. XIV at 125. Cf. Payne, 673 F.2d at 827; Carroll, 708 F.2d at 192 (appraisal system suspect, in part, because no written guidelines provided). While the district court noted, as did plaintiffs’ expert Dr. James Outtz, that the process involved subjective elements (Findings of Fact No. 189) these safeguards, together with other considerations (such as the anecdotal evidence), convinced the district court that the NLRB process did not possess the degree of subjectivity necessitating an inference of discrimination. Id. This Court has previously held that subjective factors, particularly when evaluating professional employees, may be considered by an employer in making its promotion decisions. See Wilkins v. University of Houston, 654 F.2d at 401; Page, 726 F.2d at 1046 (promotional system based upon subjective selection criteria not discriminatory per se). Given the district court’s findings, which are supported by the record, the district court did not clearly err in its conclusion that the appraisal system had not been used, “either intentionally or unintentionally, within the N.L.R.B., in a discriminatory manner against Blacks.” Conclusion of Law No. B.18. See Page, 726 F.2d at 1053.
E. Anecdotal Evidence
A plaintiff in a Title VII action may bolster a case of disparate treatment with evidence of an employer’s history of dis[1277]*1277crimination and of individual instances of discrimination. See Payne, 673 F.2d at 817. In the instant case, the district court considered evidence relating to the NLRB’s employment history and alleged individual instances of discrimination. The district court found that the NLRB’s affirmative action programs did not indicate an employer prone to discrimination, but instead one which vigorously pursued aggressive affirmative action efforts. See, e.g., Findings of Fact Nos. 5-13. On the alleged instances of individual discrimination, the district court heard testimony (which took up the major part of this lengthy trial) and specifically considered thirteen instances of individual discrimination alleged by plaintiffs. Rather than finding this evidence of a pattern or practice of discrimination, the district court concluded the opposite: that “the record as a whole could only support isolated instances of discrimination.” Finding of Fact No. 189. In many instances, this anecdotal evidence demonstrated the NLRB to be an employer which took its affirmative action duties seriously. The district court did not clearly err in its conclusions regarding the anecdotal evidence.
III. CONCLUSION OF THE CLASS CLAIM
The statistical evidence (including alleged differences between blacks and nonblacks in the percentage of employees eventually promoted out of GS levels, differences in the rate of promotions, and differences in promotions to supervisory positions) does not present the gross statistical disparity which, when considered in light of the entire record, would command a reversal of the district court’s ultimate findings of no pattern of discrimination in the NLRB’s promotional practices. Nor is the structure of the NLRB’s appraisal process or the anecdotal evidence, when combined with this statistical evidence, sufficient to support a finding of disparate treatment as a matter of law. Accordingly, we affirm the district court’s judgment on the class claim.
THE INDIVIDUAL CLAIM
Donald Lewis, a black, is the individual representing the class in this suit. He began his career with the NLRB as a field examiner in the Houston office in October 1971. Lewis was hired as a GS-9 and received favorable appraisals and regular promotions through the GS-9, 11, and 12 levels. Thus, Lewis reached the GS-12 journeyman level.
In January 1975, Lewis was rated “well qualified” for promotion to GS-13, a nonsupervisory field examiner position. Lewis received this rating sixteen months after his previous promotion to GS-12, which is the minimum waiting period for a GS-12 field examiner to receive such a “well qualified” rating.
Regional Director Louis Baldovin informed Lewis that even though he (Lewis) had been rated well qualified for the promotion, Lewis could not receive the promotion because of the NLRB’s 50/20 rule. (See supra at 1270.) Lewis knew of the 50/20 rule as early as his first year (1971-72) of employment with the NLRB. See Record Vol. XI at 42.
Lewis sought a copy of the 50/20 rule from Regional Director Baldovin in February 1975. Lewis was unable to obtain such a copy. Lewis testified that he began action toward filing an EEOC complaint in March 1975 to challenge application of the 50/20 rule. Lewis’ initial EEOC complaint claimed that the 50/20 rule discriminatorily blocked the promotional opportunities of blacks in the NLRB. (See n. 2 supra.) Plaintiff Lewis, on his individual claim, contends that after he initiated this EEOC complaint contesting the 50/20 rule, NLRB officials launched a campaign of retaliation against him. Specifically, Lewis alleged before the district court and argues here that NLRB officials retaliated against him by assigning him to a “black team,” by delaying a promised increase in salary, and by reducing his rating for promotion to the expert level GS-13 from “well qualified” in 1975 to “not ready” in 1976.
[1278]*1278II. THE MERITS OF THE INDIVIDUAL CLAIM
Section 704(a) of Title VII prohibits employers from “discriminatpng] against any of his employees ... because [the employee] has made a charge” under Title VII. 42 U.S.C. § 2000e-3(a). This Court stated the applicable order of proof in McMillan v. Rust College, Inc., 710 F.2d 1112, 1116 (5th Cir.1983):
[T]he plaintiff must first establish a prima facie case of retaliation by showing (1) that she engaged in an activity protected by Title VII; (2) that an adverse employment action occurred; and (3) that there was a causal connection between the participation in the protected activity and the adverse employment decision. Once the prima facie case is established, the burden of producing some nondiscrimihatory reason falls upon the defendant____ The employee then assumes the burden of showing that the reasons given were a pretext for retaliation.
Once the case is fully tried on the merits, as this one was, the decision of the district court as to the ultimate question of illegal discrimination must stand unless it is clearly erroneous. See United States Postal Service Board of Governors v. Aikens, 460 U.S. 711, 103 S.Ct. 1478, 1482, 75 L.Ed.2d 403 (1983); Pullman-Standard v. Swint, 456 U.S. 273, 102 S.Ct. 1781, 72 L.Ed.2d 66 (1982). On all of Lewis’ claims of retaliation, the district court was presented with conflicting evidence in which Lewis asserted a retaliatory intent and in which his supervisors denied such an intent.
Lewis contends first that shortly after he initiated his EEOC action, he was placed on a black team. Lewis believed this placement was an attempt by Regional Director Baldovin “to enveigle the senior Black in the office, Mr. Robert Penrice, to lay the groundwork for further [retaliatory] action against Mr. Lewis.” Plaintiffs-Appellants’ Brief at 28. Penrice testified that he was called by Regional Attorney Safos and told to head up a team composed of black professionals. Contradicting this testimony by Penrice and Lewis, Safos denied that he told Penrice to head up a “black team.” See Record Vol. XX at 31. Instead, Regional Director Baldovin testified that he transferred Lewis to Penrice’s team in order to resolve differences Lewis had with his previous supervisor, Dwain Irwin.
Lewis also asserts that Baldovin purposely delayed a promised Quality Within-Grade (QWI) salary increase. In January 1975, Baldovin told Lewis he would recommend Lewis for the increase within grade GS-12 since his 1975 appraisal rated him qualified for promotion to GS-13 but the 50/20 rule blocked the promotion to the new grade. The district court noted that the recommendation for the increase within-grade was an attempt to help employees whose promotions were blocked by the 50/20 rule. In March 1975, Lewis initiated his EEOC action. When Lewis did not receive the expected increase, he did not seek Baldovin for an explanation but instead complained to the NLRB’s EEO director who then contacted Baldovin. Lewis then received the recommendation for the within-grade increase shortly thereafter. Baldovin testified that he had not purposely delayed the increase. He explained that his policy was to recommend employees for within-grade increases in June of each year.19
Finally, and most seriously, Lewis contends that the ratings resulting from his 1976 appraisal (subsequent to his filing the EEOC complaint challenging the 50/20 rule in March 1975) were reduced in retaliation for the earlier EEOC action. The district [1279]*1279court found that Lewis’ 1976 appraisal was less favorable than Lewis’ previous appraisals.20 The district court, however, also noted that problems in Lewis’ performance had been noted as early as May 1975. The district court pointed out that Lewis admitted an incident noted in his 1976 appraisal (the appraisal in which Lewis received his lower recommendation) in which Lewis refused a particular assignment because Lewis contended that it was GS-13 work while he was only a GS-12, and that “[t]hey were not going to promote me anyway.” Record Vol. XI at 89. Based upon earlier appraisals and Baldovin’s own observations of Lewis, the district court concluded, Regional Director Baldovin rated Lewis not qualified for promotion to the GS-13 level in 1976. While Lewis contended that the appraisal was a product of retaliation, Baldovin testified that the lower rating was based on Lewis’ performance. See Findings of Fact No. 109 and 110.
On each of Lewis’ claims, then, the district court considered conflicting evidence. Some of this testimony substantiated Lewis’ claim of retaliation while other evidence pertained to explanations of innocent conduct by Lewis’ supervisors. The district court expressly chose the testimony of Lewis’ supervisors. It cannot be said that, in doing so, the district court clearly erred. Lewis therefore failed to establish retaliation for his previous Title VII activity.
COSTS
The district judge, upon motion by the defendants, awarded costs of $13,-289.83. Plaintiffs contend that this award was erroneous. Plaintiffs argue (1) that costs cannot be taxed against them unless the suit was frivolous, meritless or vexatious and (2) that the district court erred in concluding that this suit was frivolous.
Rule 54(d) governs the awarding of costs and establishes a presumption that “costs shall be allowed as of course to the prevailing party unless the court otherwise directs.” Fed.R.Civ.P. 54(d). Appellate review of decisions to award costs or not to do so is extremely limited; this Court has noted that the award can be overturned “only when a clear abuse of discretion is shown.” In re Nissan Antitrust Litigation, 577 F.2d 910, 918 (5th Cir.1978), cert. denied, 439 U.S. 1072, 99 S.Ct. 843, 59 L.Ed.2d 38 (1978).
Plaintiffs contend that the award of costs here is erroneous. They contend that the award of costs should be governed by the Supreme Court case in Christiansburg Garment Co. v. EEOC, 434 U.S. 412, 98 S.Ct. 694, 54 L.Ed.2d 648 (1978). In Christiansburg, the Supreme Court, in interpreting 42 U.S.C. § 2000e-5(k) of Title VII,21 held attorney’s fees should be awarded a prevailing defendant only where the suit is frivolous. This Court, however, has expressly rejected the position that Christiansburg also applies to the awarding of costs. Hill v. J.G. Penney Co., 688 F.2d 370, 375 (5th Cir.1982); Jones v. City of San Antonio, 568 F.2d 1224, 1226 (5th Cir.1978) (per curiam).
Plaintiffs, however, contend that the instant ease is distinguishable and that Christiansburg indeed does apply under the unique circumstances of this case. Plaintiffs argue that they relied on a 1978 Department of Justice memorandum that directed government attorneys to seek costs only where the suit meets the Chris-[1280]*1280Hamburg standard. This 1978 memorandum was published in a commercial reporting service. See Record Vol. IV at 58-59. This 1978 memorandum, plaintiffs argue, bound the government to seek costs only under the Christiansburg standard.
The district court did not decide whether the Department of Justice memorandum bound the NLRB to seeking costs under Christiansburg. Instead, the district court held that, assuming Christiansburg applied to the instant case, this suit was frivolous. We disagree with the district court’s conclusion that this suit was frivolous. Plaintiffs built a substantial case through statistics and anecdotal evidence. While the district court did not clearly err on the merits of this case in finding no discrimination, this suit was plainly not baseless.
We conclude that the 1978 memorandum, which served only as a policy guide for use by government attorneys, did not bind the NLRB to seek costs under the Christians-burg standard. Still, the district court’s erroneous conclusion that this suit was frivolous obviously influenced the district court in exercising its discretion to award costs to the prevailing defendant. Although one might assume that the district court would impose costs under the lower Rule 54(d) standard since it did so under the higher Christiansburg standard, we hesitate to predict such a result given the district court’s reliance on an erroneous assumption. Accordingly, we remand to the district court. In exercising its discretion under Rule 54(d), the district court may consider the extent of any reliance by the plaintiffs on the Justice Department memorandum.
CONCLUSION
We have carefully considered plaintiffs’ points of error. With the exception of the award of costs, the judgment of the district court is affirmed.
AFFIRMED AS TO THE MERITS; REMANDED AS TO AWARD OF COSTS.