dissenting.
Johnson Controls refuses to employ any woman to make batteries unless she presents medical evidence of sterility. Today this court holds the refusal lawful under Title VII. A reader of the majority opinion might be excused for thinking that the case had been fully tried — and before this court — rather than decided by a district judge on a motion for summary judgment. I think it a mistake to suppose that we can decide this case once and for all on so meager a record. It is a mistake whether we affirm, on the ground that the evidence of danger to the fetus of a woman working in an environment dense with airborne lead, combined with evidence of the difficulty of reducing the amount of lead any further, conclusively establishes the lawfulness of Johnson Controls’ policy; or reverse, with directions to enter judgment for the plaintiffs, on the ground that Title VII outlaws all fetal protection policies because all bear more heavily on female than on male workers.
Title VII forbids an employer deliberately to exclude a worker from a particular job because of the worker’s sex unless sex is a “bona fide occupational qualification reasonably necessary to the normal operation of that particular business or enterprise.” 42 U.S.C. § 2000e-2(e)(l). This'defense is central to the appeal and we should attend carefully to its scope and meaning. It is written narrowly and has been read narrowly. See, e.g., Dothard v. Rawlinson, 433 U.S. 321, 332-37, 97 S.Ct. 2720, 2728-31, 53 L.Ed.2d 786 (1977); Torres v. Wisconsin Dept. of Health & Social Services, 859 F.2d 1523, 1527-28 (7th Cir.1988). There is no useful legislative history concerning the defense, and — no doubt because the prohibition of sex discrimination was added to Title VII at the last minute— *903no reference at all to the application of the defense to sex discrimination. A narrow reading is, nevertheless, inevitable. A broad reading would gut the statute. For it is unlikely that most employment discrimination in the private sector is irrational. Few private employers discriminate without having some reason for doing so; competition tends to drive from the market firms that behave irrationally. See Becker, The Economics of Discrimination (2d ed. 1971). If the defense of bona fide occupational qualification were broadly construed — for example, to excuse all sex discrimination that the employer could show was cost-justified — very little sex discrimination in employment, as well as very little employment discrimination based on religion or national origin (forms of discrimination that, like sex discrimination but unlike discrimination based on race or color, are also excused if a bona fide occupational qualification is established), would be forbidden. Title YII’s reach would be shortened drastically.
Two courts of appeals faced with challenges under Title VII to fetal protection policies have concluded that such policies can never satisfy the stringent requirements of the occupational qualification defense. See Wright v. Olin Corp., 697 F.2d 1172, 1185 (4th Cir.1982); Hayes v. Shelby Memorial Hospital, 726 F.2d 1543, 1549 (11th Cir.1984). But this conclusion, rather than resulting in instant victory for the plaintiffs, led those courts to stitch a new defense expressly for fetal protection cases. See 697 F.2d at 1183-92; 726 F.2d at 1548-54. I am not myself deeply shocked that courts sometimes rewrite statutes to address problems that the legislators did not foresee — a notable but not isolated example being the judicial interpolation of the word “reasonable” into section 1 of the Sherman Act to prevent that atomization of society that Justice Holmes so feared. See National Society of Professional Engineers v. United States, 435 U.S. 679, 687-88, 98 S.Ct. 1355, 1363-64, 55 L.Ed.2d 637 (1978); Northern Securities Co. v. United States, 193 U.S. 197, 410, 24 S.Ct. 436, 471, 48 L.Ed. 679 (1904) (Holmes, J., dissenting). Speaking of Holmes, he wrote a splendid opinion holding that a requirement in the Massachusetts constitution of a “written vote” could be satisfied by a voting machine that involved no writing. See In re House Bill No. 1,291, 178 Mass. 605, 60 N.E. 129 (1901). And in the famous “murdering heir” case, Riggs v. Palmer, 115 N.Y. 506, 22 N.E. 188 (1889), the court in effect inserted the words “unless the legatee murdered his testator” into a statute that entitled legatees to inherit under wills executed with the proper formalities. (Notice the age of these last two cases; they were not the work of newfangled judicial activists.) Recently the First Circuit struck two words from a federal statute to make it make sense. See United States v. Colon-Ortiz, 866 F.2d 6, 11 (1st Cir.1989). There are many similar examples. I do not think judges must or should ratify absurd results by sticking doggedly to the plain meaning of statutory language.
But we do not need to bite this bullet here, because the wording of the occupational qualification provision is not so cramped that it has to be stretched to bring (some) fetal protection policies within its scope. Cf. Pime v. Loyola University, 803 F.2d 351, 356-57 (7th Cir.1986) (concurring opinion). Nor is a defensible way of stretching it to recast what is plainly a disparate treatment case — that is, a case of intentional discrimination against a protected group — as a disparate impact case, and then invoke the recent decision in which the Supreme Court expanded the “business necessity” defense. See Wards Cove Packing Co. v. Atonto, — U.S.-, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989); also Allen v. Seidman, 881 F.2d 375 (7th Cir.1989). This legerdemain is as unnecessary as it is questionable. “[Rjeasonably necessary,” one of the key terms of the occupational qualification defense, means more than just reasonable but less than absolutely necessary. On the way to concluding that the defense is unavailable in fetal protection eases the court in Wright misquoted the provision by leaving out the word “reasonably,” see 697 F.2d at 1185 n. 21, and the misquotation is faithfully repeated in Hayes, see 726 F.2d at 1549. The other key words of the de*904fense, “normal operation” (emphasis added), should dispel concern that consideration of all interests other than the employer’s interest in selling a quality product at the lowest possible price is precluded. It is possible to make batteries without considering the possible consequences for people who might be injured in the manufacturing process, just as it would be possible to make batteries with slave laborers, but neither mode of operation would be normal. To confine the occupational qualification defense to concerns with price and product quality would deny a defense to Johnson Controls even if the company excluded only pregnant women, as distinct from all women who might become pregnant, from making batteries. I do not understand the plaintiffs to be arguing that Title VII requires Johnson Controls to permit women known to be pregnant to continue working in an atmosphere dense with lead. If on the other hand a fetal protection policy that excludes women from a given job classification cannot be said to be reasonably necessary to the employer’s normal operation, I do not see why we should want to save it from condemnation under Title VII.
I have described what I conceive to be the scope of the bona fide occupational qualification defense, and its application to sex discrimination, as of the original enactment of Title VIL I must now consider the bearing of the Pregnancy Discrimination Act of 1982, which amended Title VII by defining sex discrimination to include discrimination on the basis of pregnancy. The only section amended was the definition section, 42 U.S.C. § 2000e(k). The prima facie case of sex discrimination was broadened; the defenses remained unchanged. Maybe Scherr v. Woodland School Community Consolidated District No. 50, 867 F.2d 974, 978 (7th Cir.1988), goes too far in saying that “as a definition amendment, the PDA provides no substantive rule to govern pregnancy discrimination.” For one thing, the amendment shows that the present case really is a disparate treatment case, that is, a case of intentional discrimination that can be excused only if the defendant establishes a bona fide occupational qualification; the amendment makes fertile women, the group that Johnson Controls deliberately excluded from a job classification, a group protected by Title VII. The amendment also helps us understand that the occupational qualification defense could not be merely a cost justification or reasonable relation defense, for if it were, the amendment would be ineffectual. Any employer can prove that it costs something to make an accommodation for pregnant or potentially pregnant workers and therefore that it is rational not to make the accommodation. But we already knew the defense was a narrow one.
The defense is applicable to this case and although it is of limited scope it is not the proverbial eye of a needle. In particular, the “normal operation” of a business encompasses ethical, legal, and business concerns about the effects of an employer’s activities on third parties. An employer might be validly concerned on a variety of grounds both practical and ethical with the hazards of his workplace to the children of his employees. A pregnant employee exposed to heavy concentrations of lead in the air may absorb the lead into her bloodstream and from there transmit it to her fetus through the placenta, causing, years later, mental retardation or other injury to the child. The parties agree that there is a solid medical basis for concern with fetal injury from airborne lead in the concentration found in battery plants, and this concern could in turn cause the employer to worry about being sued by injured children of his employees. Such a suit would not be preempted by workers’ compensation law, because the plaintiff would not be the worker. The employer would therefore be exposed to full common law damages, punitive as well as compensatory. The mother’s own negligence — for if she had been clearly warned of the hazard, but voluntarily became pregnant anyway and continued to work making batteries, she would be acting negligently with regard to the fetus — would not be imputed to the child and therefore would not reduce the employer’s liability. See, e.g., Collins v. Eli Lilly Co., 116 Wis.2d 166, 200 n. 14, 342 N.W.2d 37, 53 n. 14 (1984); In re Estate of Infant *905Fontaine, 128 N.H. 695, 519 A.2d 227, 230 (1986); Fabianke v. Weaver, 527 So.2d 1253, 1258 (Ala.1988). It would merely make the mother a joint tortfeasor with the employer. Moreover, she might not be negligent; the pregnancy might be involuntary, and lead can injure the fetus before the mother knows she is pregnant.
Some courts have said that to create liability, the injury to the fetus must occur after the fetus has become viable (¿ble to survive outside the mother’s body). And as I just noted, lead in the mother’s bloodstream can enter the fetus very early in the pregnancy — this presumably is the reason that Johnson Controls’ fetal protection policy is so strict. But the distinction between injury to the fetus before it becomes viable and injury after makes no sense with regard to tort liability — since the plaintiff is the child, not the fetus — and has generally and correctly been rejected. See, e.g., Renslow v. Mennonite Hospital, 67 Ill.2d 348, 10 Ill.Dec. 484, 367 N.E.2d 1250 (1977); Bergstreser v. Mitchell, 577 F.2d 22 (8th Cir.1978); Prosser and Keeton on the Law of Torts § 55, at pp. 368-69 (5th ed.1984).
Other questions concerning tort liability remain unanswered — in particular whether the standard of liability is negligence or strict liability and whether compliance with OSHA’s rules on safe levels of airborne lead is a defense. As a result it is difficult to estimate Johnson Controls’ exposure to tort liability, but it would be premature, in this age of mass-tort suits (which for example drove the asbestos industry into bankruptcy), to dismiss it as trivial. The possibility of tort suits against battery manufacturers for lead injury to the child of a female employee is not merely a theoretical one. There has already been one reported case; the plaintiff got to a jury, but lost. See Security National Bank v. Chloride, Inc., 602 F.Supp. 294 (D.Kan.1985). It is true that Johnson Controls has not yet attempted to document its liability concerns (one can understand the company’s reluctance, by drawing public attention to its exposure to tort liability, to invite suits). But in moving for summary judgment the company was not required to present all its evidence to the district judge on pain of being barred from presenting it at the trial that I believe should be held to determine whether sex is a bona fide occupational qualification for making batteries.
We'should not dismiss the concern over tort.liability as a narrow, selfish “bottom line” concern irrelevant to the purposes of Title VII. The potential cost of tort liability to Johnson Controls is an approximation of the potential cost to the children who have suffered prenatal injury from the airborne lead absorbed into their mothers’ bloodstreams. That is a social cost that Title VII does not require a company to ignore. At some point it may become large enough to affect the company’s normal method of operation and supply the ground for a bona fide occupational qualification of infertility.
A related point is that an employer might have moral qualms about endangering children or might fear the effect on his public relations. The ethical concern cannot be wholly dismissed, as could an ethical conviction that a woman’s place is in the home. We know from the controversy over abortion that many people are passionately protective of fetal welfare, and they cannot all be expected — perhaps they cannot be required — to park their passions at the company gate. That “strong [state] interest in protecting the potential life of the fetus” of which the Supreme Court spoke in Maher v. Roe, 432 U.S. 464, 478, 97 S.Ct. 2376, 2385, 53 L.Ed.2d 484 (1977), and other cases is not a judicial invention; it is the product of a groundswell of powerful emotion by a significant part of the community, and is only indirectly, although possibly substantially, in conflict with women’s workplace aspirations. Granted, in Doe v. First National Bank, 865 F.2d 864, 873 (7th Cir.1989), we assumed that the Pregnancy Discrimination Act forbids an employer to fire a woman for having an abortion, and although the point had not been argued our assumption may well have been correct. See H.Conf.Rep. No. 1786, 95th Cong., 2d Sess. 4 (1978), U.S.Code Cong. & Adm.News 1978, pp. 4749, 4766 (“no employer may, for example, fire or refuse to hire a woman simply because she has exer*906cised her right to have an abortion”). If so, the result is to place a limitation on an employer’s effort to protect fetal life. But the Pregnancy Discrimination Act affects only the prima facie case of sex discrimination. The defenses are untouched. No defense of bona fide occupational qualification was pleaded in Doe.
If the hazard to the fetus from airborne lead in the mother’s workplace is sufficiently great, if the amount of lead in the environment cannot be reduced without discontinuing the production of batteries, and if experience demonstrates that some women will become pregnant even after being clearly warned of the hazards to which the fetus would be exposed (there are many careless pregnancies, as is shown by the frequency of abortion and of illegitimate birth), I can find nothing in the text of the statute, or in its history or purpose, to prevent an employer from defending his refusal to allow fertile women to work in jobs in which they are exposed to dangerous concentrations of airborne lead on the ground that such refusal is reasonably necessary to the normal (civilized, humane, prudent, ethical) operation of his particular business. It is a matter of degree, and this we cannot assess on a summary judgment record. Of course the acceptance of the defense might be a hardship for those women who, though fertile, would not become pregnant. But hardship for the plaintiff is a possibility whenever a defense is sustained. It is no more than a possibility, here, as we shall see.
Let us not be deceived by superficial historical analogies or facile invocations of “paternalistic.” It is true that laws discriminating against women were once defended on the basis of a compelling social interest in protecting their fitness to bear and raise children, see, e.g., Muller v. Oregon, 208 U.S. 412, 421-22, 28 S.Ct. 324, 326-27, 52 L.Ed. 551 (1908), that this ground may have masked a desire to prevent women from competing with men for jobs (in any event this may have been the effect, see Landes, The Effect of State Maximum Hours Laws on the Employment of Women in 1920, 88 J.Pol.Econ. 476 (1980)), and that many modern American women resent the suggestion that women have a special responsibility for perpetuating the human race. But we do not have a discriminatory law here. A law that commands all employers in a given line of business to treat women specially cannot be equated to a decision by a firm in a competitive market to treat them specially, if only because in the latter case other firms are free to follow a different course. There is also a difference between protecting women against themselves as well as protecting children, and protecting an employer and his employees’ unborn children. A paternalistic measure is one that protects a person against himself, and insofar as Johnson Controls was motivated in adopting its fetal protection policy by concern with tort liability or adverse public relations, it was acting to protect its own interests. A fetus, moreover, is a different person (or proto-person) from its mother, and not all pregnant women fully internalize the welfare of their fetus, infant, or child. There are plenty of selfish and irresponsible parents, not all of whom are male. A fetal protection policy is less paternalistic than a maximum-hours law.
I conclude that Title VII even as amended by the Pregnancy Discrimination Act does not outlaw all fetal protection policies. Whether a particular policy is lawful is a question of fact, and since the burden of proof is on the defendant it will be the rare case where the lawfulness of such a policy can be decided on the defendant’s motion for summary judgment. This is not that rare case. Even if we accept that the amount of airborne lead in Johnson Controls’ battery-making operation is dangerous to the fetuses of female employees and that the company cannot reduce the danger further without shutting down the operation, a host of unanswered questions remains. The first concerns the feasibility of warnings as a substitute for a blanket exclusion of all fertile women. Before Johnson Controls adopted the blanket exclusion, eight women employed in the battery operation had become pregnant in three years. But we do not know what fraction of women employed in the operation this was, be*907cause — remarkably—the record does not reveal the number of women employed in the operation. And the only warning that was in effect during that period was one more likely to allay than to arouse concern. It compared the fetal hazards of airborne lead to those of cigarette smoking, and many women do not believe that smoking is highly hazardous to the fetus. The plaintiffs believe that a real “scare” warning would have deterred those eight pregnancies; maybe they are right.
We do not know what other manufacturers of batteries do about-the hazards of airborne lead to the fetus — whether they are content to rely on warnings, for example, and if so of what kind and with what effect. Apparently General Motors once had a fetal-protection policy identical to that of Johnson Controls, see Sirota, Sex Discrimination: Title VII and the Bona Fide Occupational Qualification, 55 Tex. L.Rev. 1025, 1058 n. 198 (1977); I have no idea whether it still does. The evidence of record concerning the potential hazard to the fetus through a father exposed to airborne lead is fragmentary and stale, yet if that hazard is significant the fact that Johnson Controls does nothing about it undermines the company’s argument that its fetal protection policy is motivated by concern for the fetus and reasonably necessary to the operation of the battery business. See Levin v. Delta Air Lines, Inc., 730 F.2d 994, 998-99 (5th Cir.1984). The evidentiary gap is due in part to the fact that the record on which the district judge based his grant of summary judgment was closed two and half years ago. A further consequence of the aging of the record is that we lack up-to-date information on the hazards of airborne lead to the fetus even through the mother. The record is also blank on the wages and alternative employment opportunities of the women employed in Johnson Control’s battery operation. These data would be pertinent to the plaintiffs’ ingenious although speculative argument that by depriving women of high-paying jobs, fetal protection may reduce women’s expenditures on fetal and child care, possibly harming the fetus and child as much as do the environmental hazards against which those policies are directed.
We also do not know how profitable the business of manufacturing batteries is, and therefore how vulnerable it is to fears, as yet speculative, of litigation arising from fetal damage. (The case at this stage is a tissue of speculation.) Apparently a vaccine was once withdrawn temporarily from the market because of fears of possible litigation, see H.R.Rep. No. 908, 99th Cong., 2d Sess. 6 (1986); see generally Gas-kins, Environmental Accidents: Personal Injury and Public Responsibility 161-62 (1989), and the lower the profit margin in making batteries the more plausible a concern with possible litigation becomes. The plaintiffs would have won a Pyrrhic victory if as a result of their winning this suit Johnson Controls shut down its battery operation, or if, as has happened with so many products formerly manufactured in this country, production shifted overseas and American automobile makers — which already use plenty of imported components — imported batteries made by companies that may have no regard either for fetal safety or for women’s welfare. If Johnson Controls terminated its battery operation as a result of this suit, the plaintiffs would be in the same position as if the occupational qualification defense had prevailed except that they might — which is to say, realistically, that their lawyers might — recover attorney’s fees. A “failing company” or “failing division” component of the occupational qualification defense makes at least as much sense as the counterpart provision in antitrust law.
Even on the. limited record before us, however, it is clear that the defendant’s fetal protection policy is excessively cautious in two regards: first in presuming that any woman under the age of 70 is fertile, and second in excluding a presumptively fertile woman from any job from which she might ultimately be promoted into battery making, even if her present job does not expose her to lead. Since these aspects of the policy are severable from the rest of it I do not think their deficiencies need condemn the entire policy, especially since the first is harmless because a worn-*908an too old to bear children has only to submit a letter to that effect from her doctor to be permitted to work in the battery plant. But these deficiencies do underscore the precipitancy of deciding this case in the defendant’s favor on the basis of the present record. It is not enough that an employer has some reason for adopting a policy that discriminates against women. Otherwise Los Angeles Dept. of Water & Power v. Manhart, 435 U.S. 702, 98 S.Ct. 1370, 55 L.Ed.2d 657 (1978), which struck down gendered annuity tables, would have come out the other way. Yet it would be a mistake to infer from Manhart that policies that lump all (or most) women into a class for unfavorable treatment are therefore unlawful per se. They are discriminatory, but this just means that the employer must shoulder the burden of establishing that the classification which excludes women is reasonably necessary to the normal operation of his business. That the occupational qualification defense remains available in such eases is implicit in the Court’s discussion of the defense in Manhart. See id. at 716 n. 30, 98 S.Ct. at 1379 n. 30.
The issue of the legality of fetal protection is as novel and difficult as it is contentious and the most sensible way to approach it at this early stage is on a case-by-case basis, involving careful examination of the facts as developed by the full adversary process of a trial. The record in this case is too sparse. The district judge jumped the gun. By affirming on this scanty basis we may be encouraging incautious employers to adopt fetal protection policies that could endanger the jobs of millions of women for minor gains in fetal safety and health.
But although the defendant did not present enough evidence to warrant the grant of summary judgment in its favor, there is no ground for barring it from presenting additional evidence at trial. Therefore it would be equally precipitate for us to direct the entry of judgment in the plaintiffs’ favor — something that the plaintiffs in the conclusion of their brief do not ask us to do, even though Rule 28(a)(5) of the Federal Rules of Appellate Procedure requires the appellant to place at the end of his brief “a short conclusion stating the precise relief sought.” They do argue earlier in the brief that Johnson Controls’ fetal protection policy is invalid, and if accepted this argument would obviate the need for a trial. But it is premature to accept it, as the plaintiffs themselves may have realized in formulating their request for relief. We should be as hesitant to endanger the health of children by condemning all fetal protection policies as we should be hesitant to endanger the jobs of women by placing our imprimatur on such policies. We should vacate the district court's judgment and remand for further proceedings to enable the compilation of an adequate evidentiary record.