This is an appeal from a final judgment in a sex discrimination class action in the District Court for the Southern District of New York under Title VII of the Civil Rights Act, 42 U.S.C. § 2002e-2, wherein Theresa J. Ste. Marie (Ste. Marie) is the named plaintiff and Eastern Railroad Association (ERA) and one of its components, Traffic Executive Association (TEA), are defendants. The gravamen of the complaint is that defendants discriminated against women in appointments to technical and managerial positions. The action has necessarily consumed a great deal of the time of Judge Carter over the last five years. The results of his labors are embodied in three published opinions, two of them lengthy, 72 F.R.D. 443 (1976) (class certification), 458 F. Supp. 1147 (1978) (liability), and 497 F. Supp. 800 (1980) (remedy), familiarity with which is assumed. In the first opinion, 72 F.R.D. 443, he certified Ste. Marie to represent a class consisting of all female employees of ERA. In the second opinion, 458 F. Supp. 1147, considering the case as presenting a claim of disparate treatment, he concluded that defendants had followed a policy and practice of sex discrimination during the relevant period, beginning 180 days prior to plaintiff's filing, in August 1974, of a charge of sex discrimination with the Equal Employment Opportunity Commission (EEOC). In the third opinion, 497 F. Supp. 800, he reviewed the claims of eleven employees who sought back pay or other individual relief and sustained seven. He also directed defendants to institute an elaborate program of job evaluation and classification of all positions throughout the organization. This included performance appraisal systems, publicization and explanation of performance ratings and review procedures, posting of all vacancies along with qualifications required to fill the positions, and much else, see 497 F. Supp. at 811. He awarded plaintiff's attorney a basic total fee of $140,670, plus a bonus of 10% of that amount, or $14,067, see City of Detroit v. Grinnell Corp., 495 F.2d 448, 471-72 (2 Cir. 1974), along with a fee for the services of a personnel expert, a statistical expert and his assistants, and costs. Defendants appeal from essentially all of the judge's rulings.

The number and difficulty of sex and race discrimination cases have increased to a point where it is neither practicable nor useful to write appellate opinions dealing in detail with every facet of each case. Decision turns on the particular history and practices of each employer, and opinions thus have limited precedential value. The guidelines for the determination of disparate treatment cases like this have been so clearly laid down by the Supreme Court in a series of cases beginning with McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S. Ct. 1817, 36 L. Ed. 2d 668 (1973), and continuing through Teamsters v. United States, 431 U.S. 324, 97 S. Ct. 1843, 52 L. Ed. 2d 396 (1977); Hazelwood School District v. United States, 433 U.S. 299, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977); Furnco Construction Co. v. Waters, 438 U.S. 567, 98 S. Ct. 2943, 57 L. Ed. 2d 957 (1978); Board of Trustees of Keene State College v. Sweeney, 439 U.S. 24, 99 S. Ct. 295, 58 L. Ed. 2d 216 (1979); and, most recently, Texas Department of Community Affairs v. Burdine, -- U.S. --, 101 S. Ct. 1089, 67 L. Ed. 2d 207 (1981), that there should be little need for intervention by courts of appeals. While unfortunately we find such a need to exist in this case despite the conscientious efforts of the district judge, we limit ourselves to our principal conclusions of error, without attempting to restate fully the complex and unique facts or to discuss every argument made by the parties. We are confident that when the case is reconsidered on remand, the district judge will deal with such further arguments as may still be pressed in light of the conclusions expressed in this opinion. However, a brief description of ERA's structure and past practices towards women is required.

ERA is an unincorporated association formed in 1970 to perform various traffic management and inspection services for its members, 23 northeastern railroads. It consists essentially of three component organizations: TEA, the Eastern Weighing and Inspection Bureau (EWIB) and the Railroad Perishable Inspection Agency (RPIA). TEA, with offices chiefly in New York City, analyzes proposals for rate and tariff changes submitted by member lines or shippers and makes recommendations, which are then voted upon by ERA's members. TEA also compiles and publishes tariffs and provides data processing services. EWIB monitors weight agreements and transit accounts between shippers and member lines and investigates loss and damage claims. It has 9 district offices. RPIA has offices in 32 cities. Its employees perform routine inspections for damage to shipments of perishable goods. The principal job classification in both EWIB and RPIA is that of inspector. Although there is no educational requirement for the position of EWIB inspector, an RPIA inspector must have completed two years of college education in agriculture, biology or bacteriology.

The district court found that prior to 1972 ERA hired women solely to fill secretarial positions and did not train them for responsibility in technical areas. Women were barred from participation in ERA's tuition refund program and "none were encouraged to attend Advanced Traffic School where they could have acquired basic technical skills to speed their advancement" in TEA. 458 F. Supp. at 1160. Except for positions in RPIA covered by a collective bargaining agreement, job vacancies were not advertised and promotions, initiated by management, were "based on the ad hoc subjective judgment of the appointing authority that a particular male employee had the necessary qualifications needed to fill the vacancy or to take on a new assignment." Id. at 1160-61. During this period, all ERA's managers and all inspectors were men. In 1972, however, pursuant to a conciliation agreement resulting from another Title VII charge, TEA agreed not to discriminate against women in the future. The tuition refund program was opened to women and women began attending Advanced Traffic School. Non-management positions in TEA were advertised and women were hired and promoted to fill a variety of technical positions in that organization. The significance of this history of discrimination antedating the commencement of liability and ERA's efforts to overcome it will emerge more fully below.

The overarching reason why the decision must be reversed is that the judge, having concluded, perhaps wrongly, that plaintiff had established a prima facie case, proceeded on the basis of a clearly improper standard as to the burden thereby placed on the defendant. He stated, 458 F. Supp. at 1165, that with the establishment of a prima facie case of sex discrimination "the burden shifted to the defendant to rebut plaintiff's case, and its failure to do so compels the conclusion that its policies and practices violate the Act." Continuing, he stated, 458 F. Supp. at 1165-66:

Disparate treatment of male and female employees can be sustained against a Title VII challenge only if the employer can establish business necessity justifying the disparity, or a bona fide occupational qualification reasonably necessary to the business operation. * * * A business necessity defense requires more than a mere showing that the questioned practices or policy served some legitimate managerial function, * * *. Instead, the challenged practices must be essential with no other reasonable alternative available. * * * Defendant's rebuttal evidence does not come close to meeting this standard. Accordingly, plaintiff has not only established a prima facie case of discrimination against defendant's female employees as a class, but she has met the ultimate burden of persuasion, * * * warranting the entry of judgment against defendant on the issue of liability.

This was error. It has been clear ever since McDonnell Douglas, supra, 411 U.S. at 802, 93 S. Ct. at 1824, that the burden that is shifted to the defendant by plaintiff's making out a prima facie case of disparate treatment a task described in Burdine as "not onerous", -- U.S. at --, 101 S. Ct. at 1093 is not a burden of persuading the trier of a business necessity to employ or promote a person belonging to the majority. The shifted burden is simply "to articulate some legitimate, nondiscriminatory reason for the (minority) employee's rejection." 411 U.S. at 802, 93 S. Ct. at 1824. This follows necessarily from the language and purpose of Title VII. The employer is not required to hire or promote a minority employee; rather he is forbidden to refuse to hire or promote such an employee because of race or sex. Hence he sufficiently rebuts a prima facie case by pointing to a business reason for his employment decision. By doing this he adequately negates, for the time being, the force of a plaintiff's initial showing that a qualified minority employee was denied employment or promotion which was given to a majority employee instead. The plaintiff then has the burden "to show that (the) stated reason was in fact pretext," 411 U.S. at 804, 93 S.Ct. at 1825, or, as later phrased, id. at 805, 93 S.Ct. at 1826, "that the presumptively valid reasons" for failure to hire or promote "were in fact a coverup for a racially (here sexually) discriminatory decision." Justice Powell's recent opinion in Burdine, supra,1 simply elaborates in a helpful fashion on the theme sufficiently sounded eight years ago in McDonnell Douglas from which the Court has never deviated. The district court's conclusion that defendants had followed a pattern and practice of disparate treatment must therefore fall as based on an erroneous allocation of the burden of proof.2

The district court's liability opinion is infected by other errors. The most significant flaw is its treatment of statistical data offered by plaintiff's expert which, in its words, "conclusively establishes plaintiff's claim of discrimination against women throughout defendant's work force at the managerial and technical levels," 458 F. Supp. at 1163, and of defendants' reports to the EEOC for 1974, 1975, 1976 and 1977. The statistical data showed the proportions of men and women in defendants' work force and in various categories in January, 1977 and January, 1975. Although plaintiff's expert claimed that the statistics showed discrimination in respect of salaries, the court found, 458 F. Supp. at 1156, there was no evidence of any salary disparities between men and women occupying the same position and that the gap between male and female employees in salary and salary expectations occurred because of the low representation of women in higher salaried positions. The data revealed by the statistics were as follows:

                  Clerical *        Technical *     Managerial *   Total
         Women         80   (69%)      35   (30%)       1    (1%)   116
1975     Men           31    (7%)     331   (76%)      74   (17%)   436
         Women         77   (65%)      38   (32%)       3    (3%)   118
1977     Men           24    (6%)     323   (77%)      73   (17%)   420
* Percentage of women or men in ERA's total work force.

The expert then applied the percentages of the respective sexes in the ERA's total work force to the number of jobs in each category to find what he thought a sex-neutral distribution would be. He concluded that women were grossly overrepresented at the clerical level and underrepresented at the managerial and technical levels. For 1977 the shortfalls at the managerial and technical levels were 4.09 and 9.13 standard deviations, whereas the overrepresentation at the clerical level was 14.50 standard deviations.3 Citing Castaneda v. Partida, 430 U.S. 482, 496-97, 97 S. Ct. 1272, 1281-1282, 51 L. Ed. 2d 498 (1977), and Board of Education v. Califano, 584 F.2d 576, 584 (2 Cir. 1978), aff'd sub nom., Board of Education v. Harris, 444 U.S. 130, 100 S. Ct. 363, 62 L. Ed. 2d 275 (1979); see also Hazelwood School District v. United States, supra, 433 U.S. at 311 n.17, 97 S. Ct. at 2743 n.17, the judge thought that "the grossness of the statistical disparity shown," along with other factors noted by him,4 sufficed "to establish by a fair preponderance of the evidence that defendant has (been) and is subjecting its female employees to employment discrimination because of their sex." 458 F. Supp. at 1163-64.

There can be no doubt that standard deviations of this magnitude demonstrated that the low proportions of women in the technical and managerial categories were not the result of chance. But it was a quantum leap, under the circumstances of this case, to jump from that proposition to the conclusion that such high standard deviations proved or even had any tendency to prove that defendants had committed a legally cognizable wrong. Far from being conclusive against defendants, plaintiff's statistical evidence and the EEOC reports on which it was based were totally wanting in probative value.

There is no dispute that as a rule, openings were filled by promotions from within the organization primarily on the basis of skills acquired on the job, although in some instances the taking of technical courses was either required or highly recommended. However, several members of the class, including Ste. Marie, corroborated ERA's contention that secretarial and stenographic jobs did not provide the training that could lead to advancement in other work, and that this was understood by everyone. Our review of the record shows that of the 101 "clerical" jobs existing in 1977, at least 60 were secretarial, and these were filled almost exclusively by women.5 Plaintiff's statistics were hopelessly flawed by the lumping of these secretarial jobs into the "clerical" category, since these positions did not offer the incumbents the opportunity to acquire the skills and experience that would enable them to qualify for promotion to technical and still less to managerial posts.6 There is no principle requiring an employer following a policy of promoting from within to make this applicable across the board rather than only to those employee groups whose work gives them the opportunity to acquire the skills needed for promotion. Yet plaintiff's statistics gave the same weight to failure to promote secretaries and typists to posts requiring specialized substantive knowledge and experience as they did to failure to promote women working in other clerical positions that would permit them to obtain the essential skills. This methodology failed to heed the warning in Hazelwood School District v. United States, supra, 433 U.S. at 308 n.13, 97 S. Ct. at 2742 n.13: "When special qualifications are required to fill particular jobs, comparisons to the general population (rather than to the smaller group of individuals who possess the necessary qualifications) may have little probative value." Accord, EEOC v. Radiator Specialty Co., 610 F.2d 178, 185 (4 Cir. 1979); Grano v. Department of Development of City of Columbus, 637 F.2d 1073, 1078 (6 Cir. 1980).7

An error with respect to RPIA inspectors also vitiated plaintiff's statistical presentation. No member of the class satisfied the educational requirement for the position of RPIA inspector, which has been in effect since 1940 and the propriety of which was not questioned. Since service as an inspector is a prerequisite for promotion within RPIA, the educational requirement in effect applies as well to all the RPIA inspector positions classified by plaintiff as managerial. All 96 RPIA inspector positions classified as technical and the 36 positions classified as managerial for 1977 therefore should have been excluded from plaintiff's statistical calculations.

The third respect in which plaintiff's statistics were invalid on their face lay in their failure to recognize that a large proportion of the technical and managerial positions had been filled before February 3, 1974, the date when defendants' potential liability began, 42 U.S.C. § 2000e-5(e). As in Smith v. American President Lines, Ltd., 571 F.2d 102, 108-09 (2 Cir. 1978), we need not here decide whether the time bar of § 2000e-5(e) operates as a denial of jurisdiction or merely as a statute of limitations and thus subject to tolling an issue on which the circuits are divided and the Supreme Court has not yet spoken dispositively. Under either view, it is true, as was said in United Air Lines, Inc. v. Evans, 431 U.S. 553, 558, 97 S. Ct. 1885, 1889, 52 L. Ed. 2d 571 (1977):

A discriminatory act which is not made the basis for a timely charge is the legal equivalent of a discriminatory act which occurred before the statute was passed. It may constitute relevant background evidence in a proceeding in which the status of a current practice is at issue, but separately considered, it is merely an unfortunate event in history which has no present legal consequences.

This principle is fully applicable to pattern or practice actions. In the Teamsters opinion, handed down on the same day as United Air Lines, Inc. v. Evans, the Court said that an employer might defeat a plaintiff's prima facie case by showing "that the claimed discriminatory pattern is a product of pre-Act hiring rather than unlawful post-Act discrimination." 431 U.S. at 360, 97 S.Ct. at 1867. See also Hazelwood School District v. United States, supra, 433 U.S. at 309-10, 97 S. Ct. at 2742-2743. Under United Air Lines, Inc. v. Evans, acts prior to 180 days before the charge are the equivalent of "pre-Act hiring." See Movement for Opportunity & Equality v. General Motors Corp., 622 F.2d 1235, 1245 (7 Cir. 1980); Patterson v. American Tobacco Co., 634 F.2d 744, 752 n.10 (4 Cir. 1980) (en banc), petition for cert. filed, 49 L.W. 3533 (Jan. 16, 1981); Association Against Discrimination in Employment, Inc. v. City of Bridgeport, 647 F.2d 256 at 274 (2 Cir. 1981).

As best we can discern, at least 45 of the 83 EWIB inspectors classified as technical in 1977 had been appointed to that position prior to February, 1974. All 11 EWIB inspectors at the managerial level had originally been hired prior to the period of potential liability; since inspector experience was an important qualification for promotion to those jobs, it seems fair to assume that even if class members had been hired as inspectors after February 1974, they would not have been eligible for promotion to the five management posts filled that same year or to the three additional ones filled later. Hence, these 45 technical and 11 managerial EWIB inspector positions were closed to members of the class by virtue of employment decisions that predated the onset of ERA's potential liability.8 It also appears that 41 cooper positions in RPIA and EWIB, as well as 23 of 27 executive positions in TEA had been filled prior to February, 1974. All of these posts were occupied by men.

The list could go on, but the effect on plaintiff's statistical evidence of the errors already noted is devastating. Using figures from 1977, the exclusion of 45 EWIB inspectors, 96 RPIA inspectors and 41 coopers reduces the number of "open" technical positions from 361 to 179. Of 76 managers listed by plaintiff, the exclusion of 11 EWIB and 36 RPIA inspector-managers and 23 TEA executives leaves only 6. Even if one were to use plaintiff's methodology of applying the female percentage of the entire work force to these remaining positions, a methodology which is seriously flawed by the inclusion of secretaries and typists, one would expect 21.9% of the technicals or 39, and a like percentage of the managers or 1.3, to be women. The discrepancy between these expected figures and the actual numbers of 38 women technicals and 3 women managers lacks statistical significance, particularly in a small sample of this sort. If secretaries and typists are excluded, the percentage of promotable women in ERA's work force is reduced to 12.3%, and the expected figures drop to 22 women technicals and .7 women managers.9

The district court erred also in assessing the evidence introduced by ERA to explain why no women had been engaged as RPIA inspectors until 1977, when two were hired. Defendants offered evidence that the job of an RPIA inspector is hazardous, necessitates heavy work, and involves crawling, lifting and bending. Also, particularly with respect to new hires, the work is done in the early morning, late at night and on week-ends in areas with high crime rates. Two recruiters for RPIA testified that, over the years, their enlistment efforts at agricultural colleges yielded about 210 applicants for the job of RPIA inspector, only 11 of whom were women; that two women offered such positions declined them; and that others expressed reservations about working under the conditions described.

The judge refused "to credit testimony that few women were interested in working as inspectors for RPIA based on the recruitment and interviewing of students at agricultural colleges." 458 F. Supp. at 1164. This refusal was not based on the witnesses' demeanor or on any conflicting evidence but rather on the ground that the testimony was "conclusory," this apparently meaning that the interviewers did not expect many women to be interested. Id. The judge also faulted defendants for not introducing evidence of the number of women enrolled in agricultural colleges, and added, id.: "In any event, whatever the actuality, the burden was defendant's to carry, and it has failed to do so." ERA's explanation of the dearth of women inspectors as rooted in the hazards and hardships of the job was rejected by the court as consisting of "no more than culturally biased, stereotyped concepts." Id. at 1165.

These findings reflect several errors. As we explained at the outset, it was error to assign the burden of proof to the defendants. It was also error to subject defendants to liability for failure to hire women RPIA inspectors from agricultural colleges; the class consisted solely of women already employed by ERA, and discrimination against non-members in initial hiring decisions would not have constituted discrimination against the class. Granting that evidence of disparate treatment of non-members, like pre-Act and time-barred discriminatory decisions, would be background evidence of a pattern or practice of discrimination against the class, the district court nevertheless erred in refusing to credit defendants' explanation. This was not based on what the judge characterized as "culturally biased, stereotyped concepts" of women's inability to perform hazardous or physically demanding work. The interviewers did not assert this as a justification for their failure to hire women who applied; rather they offered it to explain the lack of applications and to support their view, based on experience, that young women who had spent years in attending agricultural college would find other types of work more attractive than nocturnal prowling in railroad yards inspecting rotten food. Plaintiff offered no testimony that any women who had applied for RPIA inspectors jobs had been refused.10 The fact that defendants had finally found two women willing to accept does not disprove their assertion that they had not theretofore been able to do this, as the judge seems to have thought, 458 F. Supp. at 1164. If this acceptance had occurred before suit was brought, it would have strengthened defendants' case. Although it has less weight of this sort when it comes during the pendency of the suit, see, e. g., Jenkins v. United Gas Corp., 400 F.2d 28, 33 (5 Cir. 1968); Gamble v. Birmingham Southern R.R. Co., 514 F.2d 678, 683 (5 Cir. 1975), it still is insufficient to establish that defendants were lying when they said they had had no such success before.

The court also appears to have erred in failing to give proper weight to the problems confronting ERA in eliminating the effects of past discrimination against women in managerial positions. In February 1974, these posts were held exclusively by men and only a few positions have since opened up. In filling these posts ERA was not obliged to promote any specific number or percentage of women. The statute "does not require the employer to restructure his employment practices to maximize the number of minorities and women hired." Texas Department of Community Affairs v. Burdine, supra, -- U.S. at --, 101 S. Ct. at 1097, citing Furnco Construction Co. v. Waters, supra, 438 U.S. at 577-78, 98 S. Ct. at 2949-2950. The court seemingly prescribed a too rigorous standard when, after acknowledging that many of ERA's past discriminatory practices were no longer operative and that the case was limited to conduct occurring after February 3, 1974, it went on to state, 458 F. Supp. at 1161:

Consideration of defendant's past discriminatory practices are relevant to determine whether current prophylactic steps have eliminated all vestiges of the former gender-based discrimination.

If by this the court meant only that ERA was bound to eliminate any current discriminatory practices, that would be sound enough. However, we read the opinion as going further and insisting that defendants should have taken affirmative steps to cause the sex distribution in its higher level positions to conform to the distribution in its total work force.11 Such a view runs counter to Furnco and Burdine. As the Seventh Circuit said in Movement for Opportunity & Equality v. General Motors Corp., supra, 622 F.2d at 1245:

The residue of past discrimination is not immediately eliminated. In the instant case, it is more relevant to look not at a workforce makeup on a given day, but to the chances (the defendant) had to change its percentage of women minorities through current hiring decisions.

When these corrections are made, about all that is left of plaintiff's case apart from the seven instances of individual discrimination found in the remedies portion is ERA's history of discrimination prior to 1972 and the looseness of its promotion and hiring practices. Concerning the latter, the district court found that many openings were not advertised, and that the qualifications for those that were posted were ad hoc and home made, drawn up solely on the basis of the department head's views. The test administered to applicants for one position was also home made and the scoring bizarre. There were no written, objective criteria for evaluating a candidate's qualifications and no procedures to insure against discriminatory practices. Defendants argue that these practices are not enough to establish a pattern or practice as distinguished from individual acts of discrimination, that their record prior to 1972 is not probative of discrimination after February 1974, and that the pattern or practice aspect of plaintiff's case should be dismissed.

We would accept defendants' analysis, if we were sure that the voluminous record did not contain other probative evidence with respect to post-February 1974 discrimination. However, the joint appendix submitted by the parties runs to 10 volumes containing more than 3000 pages, while the selected exhibits compiled for our perusal number more than 80. The opinions of the district judge and the briefs of the parties necessarily were limited to emphasizing the more salient aspects of the record and it is possible, although unlikely, that there may be additional evidence of record, not yet brought to our attention, which would have a bearing on the question whether plaintiff has proved the existence of a pattern or practice. The district judge, having presided over these extended proceedings, is in a better position than an appellate panel to perform the task of collecting and evaluating the evidence in this exceedingly long record. We therefore remand this portion of the case to the district court for reconsideration on the present record, with the exception that plaintiff shall be ruled not to have sustained her burden of persuasion with respect to the RPIA inspectors.

It will nevertheless be conserving of the parties' and our own resources if we take this opportunity to set forth our views on the sufficiency of the non-statistical evidence with which we have already become familiar. This consists of the subjectivity of ERA's hiring and promotion procedures, ERA's past history of discrimination, and the seven findings of individual acts of discrimination. While evidence of subjective and discretionary promotion and transfer procedures may indeed strengthen the inference of a pattern or practice of purposeful discrimination that can be drawn from a valid statistical showing of disparities in the work force, see, e. g., Stewart v. General Motors Corp., 542 F.2d 445, 450-51 (7 Cir. 1976), cert. denied, 433 U.S. 919, 97 S. Ct. 2995, 53 L. Ed. 2d 1105 (1977); James v. Stockham Valves & Fittings Co., 559 F.2d 310, 330 (5 Cir. 1977), cert. denied, 434 U.S. 1034, 98 S. Ct. 767, 54 L. Ed. 2d 781 (1978); Grant v. Bethlehem Steel Corp., 635 F.2d 1007 (2 Cir. 1980), petition for cert. filed, 49 L.W. 3808 (Apr. 17, 1981), this is because the statistical disparities in such a case are evidence that the potential for disparate treatment created by loose procedures has been realized on a significant scale. But when, as here, relevant statistics are lacking and the probative evidence of discrimination is confined, as we are assuming, to seven individual incidents, subjective decisionmaking methods are not sufficient to establish a pattern or practice of discrimination;12 all that could be inferred is that their potential has been actualized in the same seven cases.

We also are not inclined to perceive sufficient additional support for a finding of a pattern or practice in plaintiff's evidence of discrimination in the years preceding 1972. We recognize that proof of past discrimination "might in some circumstances support the inference that such discrimination continued, particularly where relevant aspects of the decisionmaking process had undergone little change." Hazelwood School District v. United States, supra, 433 U.S. at 309 n.15, 97 S. Ct. at 2742 n.15; see Grant v. Bethlehem Steel Corp., supra, 635 F.2d at 1017. Here, however, the evidence of sex-based discrimination antedating 1974 seems to have concerned TEA exclusively, and it is undisputed that since 1972 significant progress has been made in eradicating discrimination in that organization. Although the procedures for promotion remained largely unchanged, this did not prevent sizable numbers of women from advancing through TEA's ranks where none had gone before. Under the circumstances of this case, TEA's record of discrimination prior to 1972, like the evidence of ad hoc procedures, does not significantly augment the district court's seven findings of discriminatory treatment.

The difficult question is whether assuming arguendo the accuracy of the district court's individual determinations, these seven acts, consisting of three denials of promotions to technical positions in TEA, two rejections of applications for EWIB inspector jobs, and two failures to promote women to managerial status in TEA, could be held to amount to a pattern or practice of purposeful discrimination. Teamsters instructs that plaintiff's task was to "prove more than the mere occurrence of isolated or 'accidental' or sporadic discriminatory acts"; rather she had to show that discrimination "was the company's standard operating procedure the regular rather than the unusual practice." 431 U.S. at 336, 97 S.Ct. at 1855 (footnote omitted). The Court set out Senator Humphrey's explanation:

(A) pattern or practice would be present only where the denial of rights consists of something more than an isolated, sporadic incident, but is repeated, routine, or of a generalized nature. There would be a pattern or practice if, for example, a number of companies or persons in the same industry or line of business discriminated, if a chain of motels or restaurants practiced racial discrimination throughout all or a significant part of its system, or if a company repeatedly and regularly engaged in acts prohibited by the statute.

The point is that single, insignificant, isolated acts of discrimination by a single business would not justify a finding of a pattern or practice

Id. at 336 n.16, 97 S. Ct. at 1855 n.16. Later in its opinion, the Court observed that an employer could rebut a prima facie showing of a pattern or practice by demonstrating that "during the period it is alleged to have pursued a discriminatory policy it made too few employment decisions to justify the inference that it had engaged in a regular practice of discrimination." 431 U.S. at 360, 97 S.Ct. at 1867 (footnote omitted).

On the evidence before us it could not seriously be maintained that ERA practiced sex-based discrimination across the board during the period of its potential liability; within TEA, women generally have been promoted to technical positions on an equal basis with men. This fact stamps as "isolated" the three discriminatory decisions which denied promotions to technical positions in TEA. What remains are the questions whether plaintiff has met her burden of persuasion in asserting a regular practice of discrimination in appointments to inspector positions in EWIB, and to managerial posts in TEA. As for the inspector positions, plaintiff's evidence consists solely of two applications by class members for the job of EWIB inspector, both of which the district court found were denied due to the applicants' sex.13 In one of these instances, the woman applied for a position in the Detroit office and the application was sent to New York for review, the only time this practice had ever been followed. In the other case, the director of EWIB admitted that sex was the reason for rejection.

We are of the view that these two incidents would be insufficient to support the inference of a routine or regular practice of discrimination with respect to EWIB inspector appointments. While the definition of a pattern or practice is not capable of a precise mathematical formulation, see United States v. Jacksonville Terminal Co., 451 F.2d 418, 441 (5 Cir. 1971), cert. denied, 406 U.S. 906, 92 S. Ct. 1607, 31 L. Ed. 2d 815 (1972), more than two acts will ordinarily be required. See Presseisen v. Swarthmore College, 442 F. Supp. 593, 608 (E.D.Pa.1977), aff'd mem. 582 F.2d 1275 (3 Cir. 1978); Miller v. Continental Can Co., 25 E.P.D. P 31,543 at 19,227 (S.D.Ga.1981). If there were evidence that a policy of discrimination had been adopted, perhaps two or even one confirmatory act would be enough. But the evidence with which we are acquainted is not of that ilk. The statement by the EWIB director that the applicant was being rejected for the particular position on account of her sex seems to have been merely an open admission of disparate treatment in that instance, not an announcement of a policy of general application. Similarly, the fact that the other woman's application precipitated the only instance of review by the New York headquarters, although having a tendency to show disparate treatment and the involvement of top management officials, does not establish that sex discrimination was a matter of company policy. Indeed, had sex discrimination been the routine procedure, there would have been no need for the unusual step of sending the application to the central office for additional consideration.

The district court found only two women who were denied managerial positions in TEA because of their sex. During the same period a third woman was elevated to executive status and a fourth received a promotion to an executive vacancy. It would not be necessary to give very much weight to these latter appointments, even though made after the filing of the complaint in this action, to conclude that plaintiff failed to establish by a preponderance of the evidence that TEA routinely denied women promotions to management positions. With only a handful of opportunities for bringing women into its managerial ranks, TEA raised two women and rejected two others. This hardly constitutes proof that purposeful discrimination was standard operating procedure. Barring the discovery of additional probative evidence in the record, we thus do not believe, on what has been called to our attention, that plaintiff has made out a pattern or practice of disparate treatment in managerial appointments.14

We have considered whether, despite the errors in connection with the pattern or practice holding, the court's conclusions in Part III of its remedy opinion, 497 F. Supp. at 804-11, awarding relief to seven discriminatees, could not be sustained either in whole or in part. However, we must assess all these conclusions in light of the court's statement at the beginning of Part II of the same opinion, 497 F. Supp. at 803:

At this stage of the proceeding a rebuttable presumption is operative that each member of the class was discriminated against because of her sex, and that various employment decisions during the period when unlawful discrimination was found to exist were made in pursuit of the employer's unlawful policy. The burden rests upon the employer to overcome that presumption.

This accurately stated the law, assuming that a pattern or practice had been shown, see Teamsters v. United States, supra, 431 U.S. at 360-62, 97 S. Ct. at 1867-1868; Franks v. Bowman Transportation Co., 424 U.S. 747, 772-73, 96 S. Ct. 1251, 1268, 47 L. Ed. 2d 444 (1976), but our reversal has now brought that assumption into serious question. We cannot be sufficiently certain that the determinations that seven employees were the subject of sex discrimination were not influenced by this presumption,15 and therefore deem it necessary also to remand this portion of the case to the district court for a reconsideration of the record.

The defendants' attack on the certification of the class is without merit. While we do not agree with defendants' criticisms of the district court's award of attorneys' fees and expenses, our reversal of the court's holding on the pattern or practice issue may ultimately entail a reduction. The propriety of the injunctive relief awarded by the district court will have to await the decision on remand.

The judgment is reversed and the cause remanded for further proceedings consistent with this opinion. This panel will retain jurisdiction.

Plaintiffs-appellees have filed a petition for reconsideration. This raises two points relating to the merits and one concerning costs.

One of the points concerning the merits, namely, that we erred in our holding with respect to secretaries and typists, 401, is wholly lacking in force. The record is clear that these employees did not obtain on-the-job experience that would qualify them for promotion to technical or managerial positions in TEA, EWIB or RPIA. As pointed out in fn. 7, p. 401, it was possible for secretaries to transfer laterally to nonsecretarial entry level jobs that offered the kind of training necessary for promotion in rate, tariff, data processing or inspection work. They would properly belong in the statistics when they effected such transfers not, as plaintiffs' expert treated them, before.

Plaintiffs complain with more merit about fn. 2, p. 399, in which, relying on a statement by the district judge, 458 F. Supp. 1147, 1160 (S.D.N.Y.1978), we held that the case had been tried below on a theory of disparate treatment and could not now be converted on appeal into one of disparate impact. Further examination of the record discloses that plaintiffs did make two rather limited claims of disparate impact, relating to the requirements of rate experience for promotion to managerial levels in TEA and of inspector experience for promotion within EWIB. However, if we were mistaken in some degree in footnote 2, the error was of no consequence. Considering the small number of appointments to managerial positions in EWIB and TEA that became available after February, 1974, and other facts alluded to in our opinion, particularly those at pp. 402-403, 405, 406-407, we are satisfied that a claim of disparate impact of the two policies here in question could not have been sustained.

Our opinion did not deal explicitly with the question of costs. Since the judgment was stated to have been "reversed" and the court had not "otherwise ordered", F.R.A.P. 39(a), defendants-appellants filed a bill of costs of $22,812.08, which included $17,768.06 for a voluminous joint appendix and collection of exhibits. Plaintiffs point out that defendants did not prevail on all issues and that the cause was remanded for further consideration. Hence it might be more accurate to regard the judgment as "reversed in part" in which event "costs shall be allowed only as ordered by the court", id. Ste. Marie, the named plaintiff, argues also that imposition of costs far exceeding her individual claim might inhibit further prosecution of the action. Her request is that the court direct each party to bear its own costs. We think it preferable to hold the question of costs on this appeal in abeyance until final judgment has been rendered by the district court and any appeal therefrom has been decided by us.

Except with respect to costs, the petition for rehearing is denied.