with whom GOODWIN, WALLACE, and J. BLAINE ANDERSON, Circuit Judges, join, concurring separately:

I agree that en banc resolution of a conflict, such as existed with respect to Heag-ney v. University of Washington, 642 F.2d 1157 (9th Cir.1981), and Wang v. Hoffman, 694 F.2d 1146 (9th Cir.1982), is the appropriate means of unraveling a tangle of conflicting holdings in circuit law.

On the other hand, while I agree that the mere fact that an employment practice is subjective does not shield it from attacks under the disparate impact theory, my view of the problems this case presents is different enough from that of the majority that it is best to set forth in some detail both my summary of the facts and my analysis of the law with respect to those facts. My thesis, in a nutshell, is that the disparate impact theory is designed to be applied to certain types of cases only. The majority opinion, although not holding otherwise, might unfortunately be read to suggest that the disparate treatment and disparate impact theories may be used interchangeably in any given fact situation. While this would read the opinion too broadly, it is certainly fair to say that the majority opinion provides no guidance in describing the circumstances to which each theory is applicable. This guidance is necessary to prevent the conversion of all, or substantially *1487 all, Title YII class actions into disparate impact cases.

I now turn to the facts which will be set out somewhat differently than in the majority opinion.

I.

FACTS

The five defendant canneries are located in remote and widely separated areas of Alaska. They operate for a short period each year, during the summer salmon runs, and lie vacant for the remainder of the year.

The cannery operations begin in May or June, a few weeks before the anticipated fish run, with a period known as the preseason. The companies bring in workers to assemble the canning equipment, repair winter damage to the facilities, and prepare the cannery for the onset of the canning season. Shortly before the fishing season, the cannery workers arrive. Cannery workers, who comprise the bulk of the summer work force, generally are unskilled individuals who staff the actual canning lines. These workers remain at the cannery as long as the salmon run lasts; they are guaranteed payment for a minimum number of weeks if the run is shorter than usual. When the canning is completed, the cannery workers depart and the canneries are disassembled and winterized by postseason workers.

Salmon are extremely perishable and must be processed within a short time after being caught. Because the fish runs are of short duration, cannery work involves intense and long hours. The canning process proceeds as follows. Independent fishermen catch the salmon and turn them over to companyowned boats, which transport the fish from the fishing grounds to the canneries. Cannery workers eviscerate the fish, remove the eggs, clean the fish, and place them in cans. Next, the cannery workers cook the salmon under precise time and temperature requirements established by the Food and Drug Administration (FDA) and inspect the cans to ensure that proper seals are maintained on the top, bottom, and sides.

Because of their remote location, the canneries must be completely self-contained, employing individuals in a great variety of jobs. Machinists and engineers, for example, maintain the canning equipment. Quality control personnel conduct the FDA-required inspections and record-keeping. Boat crews operate transport boats. Other tasks require, for example, cooks, carpenters, store-keepers, bookkeepers, and beach gangs for dock yard labor and construction. Because of the brevity of the salmon runs, most of the jobs are of short duration. The few permanent employees either staff the home offices in Seattle, Washington and Astoria, Oregon in the winter, or maintain the winter shipyard in Seattle.

Another consequence of the canneries’ location in remote areas is that the companies hire the necessary employees from various areas — primarily Alaska and the Pacific Northwest — and transport them to and from the canneries each year. They provide housing and mess halls at the canneries throughout the season.

Most of the cannery worker jobs, which are unskilled, are held by minorities. Most of the higher-paying jobs are held by Caucasians. The plaintiffs presented statistical evidence demonstrating the breadth of this disparity. Relying on this evidence, they challenged the following hiring practices the canneries use in filling the higher-paying jobs at issue: (1) the use of separate hiring channels and word-of-mouth recruitment for skilled workers; (2) nepotism; (8) rehire policies; and (4) the lack of objective job qualifications. They also alleged racial discrimination in the canneries’ messing and housing practices.

The district court evaluated all of the practices under the disparate treatment model; it found for the defendants, holding that they had shown nondiscriminatory motivations for these practices. It also evaluated some of the practices, those it characterized as “objective,” under the disparate impact model; it found for the defendants under this analysis also.

*1488 The panel to which this case was assigned agreed with the district court that disparate impact analysis should be applied only to “objective” factors. Its conclusion was based on Heagney v. University of Washington, 642 F.2d 1157 (9th Cir.1981), but conflicted with Wang v. Hoffman, 694 F.2d 1146 (9th Cir.1982). See Atonio v. Wards Cove Packing Co., 763 F.2d 1120 , 1132 & n. 6 (9th Cir.1985).

As already mentioned, we granted en banc review to address the circumstances under which it is appropriate to employ the disparate impact analysis. Part II of this opinion sets forth an analytic framework for determining when the disparate impact approach should be used. Part III applies that framework to the facts of this case.

II.

ANALYTIC FRAMEWORK

The relevant section of Title VII, 42 U.S.C. § 2000e-2(a)(2), provides:

The Supreme Court’s interpretation of this provision has identified two separate theories of recovery: disparate treatment and disparate impact. Put briefly, a plaintiff alleging disparate treatment must demonstrate intentional discrimination. See, e.g., 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). A disparate impact claim, on the other hand, does not require proof of discriminatory intent. Instead, it attacks “employment practices that are facially neutral in their treatment of different groups but that in fact fall more harshly on one group than another.” Id. at 336 n. 15.

The Supreme Court has not clearly articulated the types of cases to which each of these theories should be applied. In Teamsters, for example, the Court said that “[ejither theory may, of course, be applied to a particular set of facts.” Id. One could conclude from this comment that both theories were applicable to all Title VII claims without regard to their specific nature.

This conclusion, however, is plainly inconsistent with the Supreme Court’s disposition of Fumco Constmction Corp. v. Waters, 438 U.S. 567 , 98 S.Ct. 2943 , 57 L.Ed.2d 957 (1978). In that case, the Supreme Court expressly refused to apply disparate impact analysis. The plaintiffs were individual bricklayers who were not hired because they applied at the jobsite, rather than through the regular application process. The Supreme Court’s explanation consisted of a footnote stating that the case was not similar to Griggs v. Duke Power Co., 401 U.S. 424 , 91 S.Ct. 849 , 28 L.Ed.2d 158 (1971) (evaluating standardized tests under disparate impact analysis), Do-thard v. Rawlinson, 433 U.S. 321 , 97 S.Ct. 2720 , 53 L.Ed.2d 786 (1977) (evaluating height and weight requirements under disparate impact analysis), or Teamsters (a class action disparate treatment case). See Fumco, 438 U.S. at 575 n. 7, 98 S.Ct. at 2948 -49 n. 7. 1

It should not be surprising that the lower courts have employed different explanations of this footnote in Fumco. Two basic explanations have emerged, one represented by Wang and the other by Heagney . The Heagney approach restricts the disparate impact analysis to objective practices; *1489 the Wang approach applies it to all practices. 2 I think both of these approaches ignore how the alleged practice functions. As a consequence, one is too broad and the other too narrow.

Moreover, the distinction between “objective” and “subjective” employment practices or criteria is not as clear as these cases suggest. A requirement, for example, that an applicant pass a qualifications test is “objective.” On the other hand, hiring on the basis of good looks and appearance is by no means entirely “subjective.” Specific aspects of these two criteria can be identified and to the extent so identified become “objective.” Only an employment practice resting entirely on personal whim and caprice can be said to be wholly “subjective.” In short, “subjective” and “objective” are only the extremes of a continuum, like night and day. I believe they are inappropriate tools for defining the bounds of disparate impact and disparate treatment analysis. Moreover, even “subjective” practices, as the majority points out, have the same capacity to cloak discrimination that in Griggs led the Supreme Court to create disparate impact analysis.

I think the key to understanding the proper spheres of disparate impact and disparate treatment analysis is found in the nature of the claims of discrimination. A brief recapitulation of the nature of the two forms of analysis demonstrates this point. To establish a prima facie disparate impact case requires that the practice be identified, that there exists an impact adverse to a protected class, and that the practice caused the adverse impact.

Obviously, the burden of establishing this prima facie case will preclude certain claims from receiving disparate impact analysis. For example, the requirement that the plaintiffs identify a specific practice prevents plaintiffs from “launchpng] a wide ranging attack on the cumulative effect of a company's employment practices.” Spaulding v. University of Washington, 740 F.2d 686, 707 (9th Cir.) (quoting Pouncy v. Prudential Ins. Co. of Am., 668 F.2d 795,800 (5th Cir.1982)), cert. denied, 469 U.S. 1036 , 105 S.Ct. 511 , 83 L.Ed.2d 401 (1984). But cf. Griffin v. Carlin, 755 F.2d 1516, 1523-25 (11th Cir.1985) (applying disparate impact analysis to the end result of a hiring process, without requiring the plaintiffs to articulate which specific practices caused the impact in question). Absent this requirement, the disparate impact test would put on employers the burden of demonstrating the business necessity of each facet of their employment decisions, even if the plaintiffs could demonstrate no disparate impact caused by some of those facets. See Pouncy, 668 F.2d at 801 . Accordingly, the analysis requires the plaintiff to identify some specific practice; the defendant must show the business necessity of that specific practice.

The requirement of causation also prevents disparate impact analysis of certain claims. For example, a plaintiff’s class consisting of children cannot state a cause of action against an employer merely because his recruiting practices designed to obtain quarry workers overlooked children. No significant number of children are qualified to be quarry workers. Because there are not a significant number of children so *1490 qualified, the employer’s practices in recruiting quarry workers cannot be said to have caused any impact on the children. At a minimum, then, the causation element requires demonstration by the plaintiff that significant numbers of the plaintiff class are qualified for the job. See, e.g., Segar v. Smith, 738 F.2d 1249, 1274 (D.C.Cir.1984), cert. denied, 471 U.S. 1115 , 105 S.Ct. 2357 , 86 L.Ed.2d 258 (1985); Grant v. Bethlehem Steel Corp., 635 F.2d 1007, 1019 (2d Cir. 1980) (noting that some members of the plaintiff class were clearly qualified, despite the employers’ protestations to the contrary), cert. denied, 452 U.S. 940 , 101 S.Ct. 3083 , 69 L.Ed.2d 954 (1981). 3

Once the plaintiff has established a pri-ma facie case, the employer must either attack one of the three elements of the prima facie case or demonstrate that the practice is a “business necessity.” The latter can be shown only when the practice is job-related and serves to help identify the qualities necessary to perform the work satisfactorily. See, e.g., Dothard v. Rawlinson, 433 U.S. 321 , 332 n. 14, 97 S.Ct. 2720 , 2728 n. 14, 53 L.Ed.2d 786 (1977); Griggs v. Duke Power Co., 401 U.S. 424, 431 , 91 S.Ct. 849, 853 , 28 L.Ed.2d 158 (1971).

The disparate treatment structure is quite different. There the prima facie case typically requires that the aggrieved show (1) that he is a member of a protected class, (2) that he applied, (3) that he was rejected, and (4) that after the rejection the position remained open and applicants having qualifications similar to the aggrieved’s continued to be accepted. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 , 93 S.Ct. 1817, 1824 , 36 L.Ed.2d 668 (1973). The burden then is on the employer to show that a non-discriminatory reason explains his conduct. See id. at 802-03 , 93 S.Ct. at 1824-25 . Thereafter, the aggrieved may attempt to show that the proffered explanation is pretextual. See id. at 804 , 93 S.Ct. at 1825 . The ultimate burden of persuasion remains on the aggrieved to show discriminatory treatment. Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 253 , 101 S.Ct. 1089, 1093-94 , 67 L.Ed.2d 207 (1981).

The Supreme Court cases to date have applied disparate impact analysis only to practices akin to counting, weighing, and measuring, an even narrower limitation than the “objective”/“subjective” distinction some courts have adopted. I think the appropriate distinction can be more accurately delineated. As I see it, disparate impact analysis should be applied whenever the plaintiff claims that the employer has articulated an unnecessary practice that makes the plaintiff’s true qualifications irrelevant. This differs from a treatment case, in which the plaintiff claims that, knowing the plaintiff’s qualifications, the employer refused to hire him because of race or some other impermissible characteristic. A showing of discriminatory intent is extremely difficult, if not impossible, when an employer asserts that he did not hire an individual because of a facially neutral requirement. Faced with this reality the Court in Griggs held that employers must justify such requirements under the business necessity test.

The crucial issue in any Title VII case is into which category the employer’s alleged wrong properly fits. Has the employer allegedly failed by reason of some facially neutral employment practice to ascertain the qualifications of a protected class, or has the employer ignored the known qualifications for a discriminatory reason? The nature of the wrong as pleaded and proved *1491 determines the nature and extent of the plaintiffs burden. Because it would be futile in an impact case to require the plaintiff to show discriminatory intent, the plaintiffs burden principally is one of showing the “impact” of the practice. Proof of the “impact” goes far toward establishing a failure to consider the qualifications of a substantial number of the protected class. At that point the employer’s response logically can only be that the practice serves to ascertain a relevant job-related qualification; that is, the practice rests on business necessity.

This burden of showing a business necessity has no place if the plaintiff’s grievance is that his qualifications, although available to and known by the defendant employer, have been ignored because of a discriminatory motive. To treat this as an impact case rather than a treatment case would relieve the plaintiff of the burden of establishing a discriminatory intent and impose on the defendant the burden of demonstrating that what he did was done because of business necessity. In the context of a treatment case, this would amount to imposing the burden on the defendant to prove that he did not discriminate.

Thus, it is necessary to determine from the pleadings and the evidence the nature of each claim the plaintiff makes. Although it is true that neither impact nor treatment analysis can be tied irrevocably to a specific category of practices, it is also true that they properly cannot be employed interchangeably. It follows that in this case each claim must be analyzed to determine which type of analysis, impact or treatment, is proper. An employee, alleging only that the employer’s failure to hire him is based on race or religion, cannot force the employer to prove that his failure was due to business necessity. This remains true even if the plaintiff shows that others of plaintiff’s race or religion also had not been hired. The employee has alleged a treatment case and the burdens are allocated as McDonnell Douglas and Burdine indicate. On the other hand, such an allocation is entirely inappropriate where the allegation is that the test employed by the employer disqualifies all applicants other than Protestants. This pleads an impact case.

Complications arise when the practices' lend themselves to being alleged as the basis of either a treatment or impact case. Equally complicated are situations in which multiple practices are employed and some properly suggest impact analysis while others treatment analysis. In such situations, a court should evaluate each practice separately, applying the appropriate analysis to each practice. Guided by this analysis, I now proceed to examine the district court’s treatment of the plaintiff’s claims in this case. 4

*1492 III.

APPLICATION TO THIS CASE

A. Separate Hiring Channels and Word-of-Mouth Recruitment

The first practice the plaintiffs challenge is the use of separate hiring channels and word-of-mouth recruitment for cannery workers and for the skilled at-issue jobs. *1493 The use of separate hiring channels can insulate an employer’s decisionmaking process from any need to consider the qualifications of unwanted minorities. Accordingly, disparate impact analysis of this claim is appropriate. 5

But this does not mean that Atonio’s claim must prevail. As part of his prima facie case, he must establish causation. In turn, that element requires proof that a substantial number of the class possess the qualifications legitimately required for the skilled jobs. The district court did not make any findings on this point. Because the record is unclear, I would remand for further factfinding on this point. See Icicle Seafoods, Inc. v. Worthington, — U.S. -, 106 S.Ct. 1527, 1530 , 89 L.Ed.2d 739 (1986). For each job that the district court finds a substantial number of qualified plaintiffs, the district court must evaluate the business necessity of separate hiring channels.

B. Nepotism

The second hiring practice the employees challenge is nepotism. The district court subjected this claim to impact analysis pursuant to our decision in Bonilla v. Oakland Scavenger Co., 697 F.2d 1297, 1303-04 (9th Cir.1983), cert. denied, 467 U.S. 1251 , 104 S.Ct. 3533 , 82 L.Ed.2d 838 (1984). It rejected the claim, finding that the individuals were hired because of their abilities rather than their relation to the employers. Excerpt of Record (E.R.) at 324-25. I might construe this as a finding that the canneries had no practice of nepotism, apart from their admitted practice of word-of-mouth recruitment. If this were so, the plaintiffs’ challenge would fail. Because the appropriate legal standard was less than clear at the time the district court considered this case, I would remand this claim back to that court for further consideration.

C. Rehire Policies

The third practice the employees challenge is the rehire policies of the employers. Like the practices discussed above, rehire policies insulate the employer from the need to consider the applications of possibly qualified minorities. The district court properly applied disparate impact analysis to this practice, but rejected the employees’ challenge because it found the practice was justified by business necessity, viz. the short season and the dangers of the industry. E.R. at 334. Because this finding is not clearly erroneous, I would affirm the district court’s disposition of this claim without addressing other aspects of it.

D. Lack of Objective Employment Criteria

Next, the employees challenge the employers’ lack of objective employment criteria. The district court found as a fact that the employers did have objective criteria. The defendants’ pretrial order listed a number of qualifications assertedly necessary for the jobs in question. After hearing evidence, the court explicitly found that these qualifications were “reasonably required for successful performance.” E.R. at 299. Although some evidence in the record suggests that these qualifications were not applied evenhandedly, discrimination in application raises a treatment claim. It is only the choice of qualifications that is subject to disparate impact analysis. I cannot say that the district court’s decision *1494 was clearly erroneous. Accordingly, I would affirm its disposition of this claim.

E. Housing and Messing Practices

Finally, the employees allege racial discrimination in the canneries’ housing and messing practices. I do not think this claim is properly susceptible to disparate impact analysis. In no way do these practices enable an employer to reject prospective minority employees without considering their qualifications. The only Title VII challenge to these practices can be under the disparate treatment theory. The district court’s rejection of the claim on that theory, E.R. at 336-37, was not clearly erroneous. Accordingly, I would affirm the district court’s treatment of this claim.

In summary, I would affirm the district court’s dismissal of the plaintiffs’ claims regarding rehire policies, subjective employment criteria, and racial discrimination in housing and messing practices. I would reverse the district court’s dismissal of the separate hiring channels and nepotism claims and would remand for further fact-finding.

. Comprehension of the court’s treatment of the impact claim in Fumco is complicated by Justice Marshall’s explanation. He argues that the Court’s rejection of the impact claim was merely an affirmance of the circuit court’s affirmance of the district court’s rejection of that claim on the merits. 438 U.S. at 584-85 , 98 S.Ct. at 2953 (Marshall, J., concurring in part, dissenting in part.) Because this explanation is not consistent with the explanation of the Court’s own opinion, I refuse to rely on it.

. The decisions in other circuits in fact reflect a more complicated situation, with a variety of different positions. It is fair to say, however, that some courts apply disparate impact analysis only to practices closely akin to the counting, measuring, and weighing evident from the existing Supreme Court cases. See, e.g., Carroll v. Sears, Roebuck & Co., 708 F.2d 183, 188-89 (5th Cir.1983) (Wisdom, J.) (refusing to apply disparate impact analysis to claims of discrimination in training, promotion, and classification of employees); Harris v. Ford Motor Co., 651 F.2d 609, 611 (8th Cir.1981) (per curiam) (refusing to apply disparate impact analysis to system allowing firing based on evaluations of supervisors). Other courts apply disparate impact analysis to any identifiable practice whatsoever. See, e.g., Rowe v. Cleveland Pneumatic Co., 690 F.2d 88, 92-93 (6th Cir.1982) (per curiam) (applying disparate impact analysis to system allowing rehiring based on opinions of foremen); Clark v. Chrysler Corp., 673 F.2d 921, 927 (7th Cir.) (applying disparate impact analysis to word-of-mouth recruitment and discriminatory selection of hiring channel), cert. denied, 459 U.S. 873 , 103 S.Ct. 161 , 74 L.Ed.2d 134 (1982).

. I do not mean to say that plaintiffs must introduce statistical proof based on qualifications of applicants who have been rejected for the job. Obviously, the applicant pool itself could fail to represent adequately the number of qualified minorities because of discriminatory recruitment practices. See, e.g., Dothard v. Rawlinson, 433 U.S. 321, 330 , 97 S.Ct. 2720, 2727 , 53 L.Ed.2d 786 (1977). Those discriminatory recruitment practices themselves are subject to disparate impact analysis. But it is important to remember that a prima facie case that the recruitment practices in question have caused a disparate impact requires demonstration of a significant number of qualified persons overlooked because of the challenged practices.

. I acknowledge that this position has not been articulated in the decisions of other courts that have examined similar questions. A brief survey of the law in other circuits reveals, however, that most of the decisions in this area are consistent with the approach I suggest.

The Second Circuit has applied disparate impact analysis to employment systems that relied on subjective employee evaluations. Zahorik v. Cornett Univ., 729 F.2d 85, 95-96 (2d Cir.1984). Under my approach, such decisions would often be subject to the disparate impact analysis.

The Third Circuit applied disparate impact to invalidate a test that partially based promotions on administrative skills. In that case, the employer had a practice of assigning whites to jobs that developed the administrative skills tested for by the exam. Accordingly, reliance on the administrative skills was improper. See Wilmore v. City of Wilmington, 699 F.2d 667, 675 (3d Cir. 1983).

None of the Fourth Circuit decisions commonly cited in this area seems to have dealt specifically with the objective/subjective distinction. For instance, in EEOC v. Federal Reserve Bank, 698 F.2d 633 , 638-39 (4th Cir. 1983), rev’d on other grounds sub nom. Cooper v. Federal Reserve Bank, 467 U.S. 867 , 104 S.Ct. 2794 , 81 L.Ed.2d 718 (1984), the court flatly stated that disparate impact analysis could be applied only to objective practices. In that case, however, the plaintiffs apparently identified no specific practice; instead, they seem to have been challenging the entire employment process. I would reach the same result, refusing to apply disparate impact unless the plaintiffs can identify a specific practice that causes a disparate impact. Similarly, Pope v. City of Hickory, 679 F.2d 20 (4th Cir. 1982), was a disparate treatment case; the plaintiffs alleged discrimination in general, not that it was implemented through some specific practice. Brown v. Gaston County Dyeing Machine Co., 457 F.2d 1377 (4th Cir,), *1492 cert. denied, 409 U.S. 982 , 93 S.Ct. 319 , 34 L.Ed.2d 246 (1972), failed to distinguish between the impact and treatment analysis at all. Robinson v. Lorillard Corp., 444 F.2d 791 (4th Cir.), cert. dismissed pursuant to Sup.CtR. 60, 404 U.S. 1006 , 92 S.Ct. 573 , 30 L.Ed.2d 655 (1971), is actually precedent for application of disparate impact analysis to more subjective systems, despite the flat statement in EEOC. In Robinson , the Fourth Circuit applied disparate impact to use of a seniority system that was at least partially subjective.

The decisions in the Fifth Circuit display a similar lack of resolution in drawing a line between objective and subjective practices. Several panels of that circuit have thought that the law of the circuit precluded application of the disparate impact analysis to subjective factors, relying on Pouncy v. Prudential Insurance Co. of America, 668 F.2d 795 (5th Cir. 1982). See Vuyanich v. Republic Natl Bank, 723 F.2d 1195 ,1201-02 (5th Cir.), cert. denied, 469 U.S. 1073 , 105 S.Ct. 567 , 83 L.Ed.2d 507 (1984); Carroll v. Sears, Roebuck & Co., 708 F.2d 183, 188-89 (5th Cir.1983) (Wisdom, J.); Pegues v. Mississippi State Employment Serv., 699 F.2d 760 , 764 (5th Cir.), cert. denied, 464 U.S. 991 , 104 S.Ct. 482 , 78 L.Ed.2d 679 (1983). But at least one recent Fifth Circuit panel noted Pouncy and went on to apply disparate impact analysis to a system that based promotions on the subjective evaluations of foremen. See Page v. United States Indus., 726 F.2d 1038 , 1045-46 (5th Cir.1984). The clarity of the ostensible rule of Pouncy is also not evident from that opinion itself. In fact, the opinion had alternative holdings: first, that the plaintiffs had not established that the practices caused the impact; and, second, that the practice was not susceptible to the disparate impact analysis because of its subjectivity. 668 F.2d at 800-01 . I also note that in none of the Fifth Circuit cases following Pouncy would plaintiffs clearly have prevailed under my disparate impact analysis anyway. See Vuyanich, 723 F.2d at 1201-02 (plaintiff apparently failed to identify a specific practice); Carroll, 708 F.2d at 188-90 (apparently the plaintiffs failed to show causation); Pegues, 699 F.2d at 764-65 (practice not by an employer, but by a state employee commission).

In the Sixth Circuit, disparate impact analysis has been applied in cases challenging rehiring based on unguided opinions of foremen. See Rowe v. Cleveland Pneumatic Co., 690 F.2d 88, 92-93 (6th Cir. 1982) (per curiam).

The Seventh Circuit, in a case strikingly similar to this one, applied disparate impact analysis, as I do here, to word-of-mouth recruitment and selection of hiring channels. See Clark v. Chrysler Corp., 673 F.2d 921, 927 (7th Cir.), cert. denied, 459 U.S. 873 , 103 S.Ct. 161 , 74 L.Ed.2d 134 (1982).

In the Eighth Circuit, I do find cases that are not reconcilable with my approach. That circuit has maintained a firm refusal to apply disparate impact analysis to what it characterizes as "subjective” practices. See, e.g., Gilbert v. Little Rock, 722 F.2d 1390 (8th Cir.1983) (applying treatment analysis to a system relying on individual discretion), cert. denied, 466 U.S. 972 (1984); Talley v. United States Postal Serv., 720 F.2d 505, 506-07 (8th Cir.1983) (refusing to apply impact analysis), cert. denied, 466 U.S. 952 , 104 S.Ct. 2155 , 80 L.Ed.2d 541 (1984); Harris v. Ford Motor Co., 651 F.2d 609 (8th Cir.1981) (per curiam) (same). For the reasons articulated in the text, I think these cases are incorrect. I note that this footnote demonstrates that my approach is consistent with the great majority of existing authority.

The Tenth Circuit has uniformly applied disparate impact analysis to practices that use subjectivity to cloak discrimination. See, e.g., Hawkins v. Bounds, 152 F.2d 500, 503 (10th Cir. 1985); Lasso v. Woodmen of the World Life Ins. Co., 741 F.2d 1241, 1245 (10th Cir.1984), cert. denied, 471 U.S. 1099 , 105 S.Ct. 2320 , 85 L.Ed.2d 839 (1985); Coe v. Yellow Freight Sys., 646 F.2d 444, 450-51 (10th Cir.1981) (dicta); Williams v. Colorado Springs, Colo. School Dist. No. 11, 641 F.2d 835 (10th Cir.1981).

I have already noted the inconsistency of one recent Eleventh Circuit decision with my opinion. See Griffin v. Carlin, 755 F.2d 1516, 1523-25 (11th Cir. 1985) (applying disparate impact analysis to the end result of a hiring process without requiring the plaintiffs to identify a particular practice). That disagreement as to the requirements of the prima facie case does not extend, however, to the scope of the impact analysis itself. I would apply impact analysis to the facts of Griffin , only reaching a different result.

Finally, the D.C. Circuit has recently articulated a complicated position, not completely in accord with either of the common positions exhibited in the other circuits. See Segar v. Smith, 738 F.2d 1249, 1270-72 (D.C.Cir.1984), cert. denied, 471 U.S. 1115 , 105 S.Ct. 2357 , 86 L.Ed.2d 258 (1985). In that opinion, the panel discussed the following scenario. After a plaintiff establishes a prima facie treatment case, defendants frequently advance an employment practice as a legitimate reason for their hiring decisions. According to the Segar panel, the employers' articulation of that practice as a defense to the treatment case establishes a prima facie impact case against the practice in question. Accordingly, the defendants must defend the practice under the business necessity test required by disparate impact analysis.

. I recognize that this claim is quite similar to the claim presented in Fumco Construction Corp. v. Waters, 438 U.S. 567 , 98 S.Ct. 2943 , 57 L.Ed.2d 957 (1978), a claim to which the Court refused to apply disparate impact analysis, id. at 575 & n. 7, 98 S.Ct. at 2948 -49 & n. 7. In that case, the Court emphasized the "importance of selecting people whose capability has been demonstrated to defendant.” Id. at 574 , 98 S.Ct. at 2948 (quoting the lower court opinion). If this were treated as a job qualification, under my analysis the impact analysis would apply, but the plaintiffs would have failed to establish a prima facie case because they were not qualified. Most importantly, however, the Fumco footnote is just not specific enough to resolve the question before us. I do not think it is useful to search at length for an explanation for the Fumco result the Court declined to give us.