joined by MANION, Circuit Judge, concurring and dissenting.
These cases present the court with difficult and important questions concerning the application of Title VII of the Civil Rights Act of 1964 to sexual harassment by supervisory employees. I agree with the basic approach to these questions, and most of the specific answers to them, in Judge Flaum’s concurring opinion. Our most substantial difference concerns the employer’s liability, in “quid pro quo” sexual harassment cases, for what I shall call “noncompany acts”; these are illustrated by a supervisor’s threat that is not followed up by termination, demotion, or other acts that change the contractual relation between the victimized employee and the employer. I shall explain this difference between us after setting forth a general approach that differs somewhat from his.
When the Supreme Court in Meritor Savings Bank, FSB v. Vinson, 477 U.S. 57, 72, 106 S.Ct. 2399, 2408, 91 L.Ed.2d 49 (1986), said that “Congress wanted courts to look to agency principles” to determine the liability of employers sued under Title VII for sexual harassment of one employee by another, the Court was not ruling that Title VII incorporates the American Law Institute’s Restatement of Agency. Obvious as this point should be to any reader of the Supreme Court’s opinion' — for the opinion states that “common-law [agency] principles may not be transferable in all their particulars to Title VII,” 477 U.S. at 72, 106 S.Ct. at 2408 — it hasn’t sunk in all the way. One of the briefs in the cases before us actually states, “Since Congress placed a comma after the first clause of § 219(2)(d) [of the Restatement;].... ” The courts do not make this mistake; but the Restatement has become their template for analyzing the issue of employers’ liability for harassment by supervisors. This is the wrong approach, and I am pleased that Judge Flaum does not follow it.
In saying that it is wrong to rely on the Restatement, I do not mean that it is wrong to bring agency principles into the application of Title VII. That is inescapable. Title VII creates a statutory tort of employment discrimination but imposes liability only on the “employer”; what is to count as the employer’s discriminatory act, and so warrant a remedy under Title VII, is a question of agency that the federal courts must by default answer as a matter of federal common law. Often courts create a federal common law principle simply by borrowing a part of the law of the state in which the legal dispute arose. That would not be a sound approach here, and a majority of the judges of this court reject it; nor is it the tenor of the Supreme Court’s reference in Meritor to the judicial role: “Congress wanted courts to look to agency principles.” Given that the employer is the only person made liable by the statute, the question of agency is central to its administration, so that to defer to state *507law could introduce striking geographical disuniformities and indeed would empower a state to nullify Title VII.
It is true that the Supreme Court recently repudiated a comprehensive federal judicial power to create common law, but it was speaking of federal common law in “its strictest sense,” Atherton v. FDIC, — U.S. —, —, 117 S.Ct. 666, 670, 136 L.Ed.2d 656 (1997), that is, of legal doctrine that could not be construed as interpretive. In Atherton it was a standard of care for directors of all financial institutions and had no more warrant than the fact that the institution was federally chartered or insured. Deciding what agency principles shall govern liability under a liability-creating statute such as Title VII is not free-wheeling common-law rule-making; it is filling a statutory gap, a standard office of interpretation. There is no novelty in formulating federal principles of agency law in interpreting federal statutes that are silent on agency. See, e.g., Central States Trucking Co. v. J.R. Simplot Co., 965 F.2d 431 (7th Cir.1992); United States v. Balistrieri, 981 F.2d 916, 930 (7th Cir.1992). I have not found a ease which suggests that liability under Title VII should depend on the agency law of the state where the dispute arose. The parties to the cases before us have not suggested that we take such an approach, and Judge Manion’s opinion marshals compelling authority against it.
Both cases before us arose in Illinois, but the parties have not enlightened us about the Illinois law of agency. In fact Illinois courts almost always refuse to hold employers liable for the sexual misconduct of their employees, including their supervisory employees. See Deloney v. Board of Education, 281 Ill. App.3d 775, 217 Ill.Dec. 123, 128-29, 666 N.E.2d 792, 797-98 (1996) (truant officer who molested 16-year-old student); Randi F. v. High Ridge YMCA, 170 Ill.App.3d 962, 120 Ill.Dec. 784, 788, 524 N.E.2d 966, 970 (1988) (sexual assault by employee of day-care center on 3-year-old); Webb v. Jewel Cos., Inc., 137 Ill.App.3d 1004, 92 Ill.Dec. 598, 485 N.E.2d 409 (1985) (sexual assault by security officer on suspected shoplifter); Hoover v. University of Chicago Hospitals, 51 Ill. App.3d 263, 9 Ill.Dec. 414, 366 N.E.2d 925 (1977) (sexual assault by doctor on patient); Dockter v. Rudolf Wolff Futures, Inc., 684 F.Supp. 532, 535-36 (N.D.Ill.1988) (sexual harassment by manager) (interpreting Illinois law), aff’d, 913 F.2d 456 (7th Cir.1990). St. Paul Fire & Marine Ins. Co. v. Downs, 247 Ill.App.3d 382, 187 Ill.Dec. 130, 617 N.E.2d 338 (1993), found sexual molestation of a patient by a psychiatrist to be arguably within the scope of the psychiatrist’s employment, but based this on the patient’s psychological dependence on the psychiatrist, distinguishing the other cases that I have cited. Id. at 136, 617 N.E.2d at 344. Downs belongs to the class of eases, illustrated also by sexual molestation by police officers, as in our recent West v. Waymire, 114 F.3d 646, 649 (7th Cir.1997), in which the employer has entrusted the employee with unique power to molest or otherwise oppress subordinates or members of the public; the distinction between these cases and the ordinary case of sexual harassment by a supervisory employee is' stressed in Deloney v. Board of Education, supra, 281 Ill.App.3d 775, 217 Ill.Dec. at 130 n. 5, 666 N.E.2d at 799 n. 5. In Deal v. Byford, 127 Ill.2d 192, 130 Ill.Dec. 200, 537 N.E.2d 267 (1989), scope of employment was admitted, hence not an issue, and anyway the employee was clearly acting in furtherance of the employer’s interests. See id. at 203, 206-07, 537 N.E.2d at 270, 272-73. And Pyne v. Witmer, 129 Ill.2d 351, 135 Ill.Dec. 557, 543 N.E.2d 1304 (1989), did not involve an intentional tort.
Green Hills Country Club v. Illinois Human Rights Comm’n, 162 Ill.App.3d 216, 113 Ill.Dec. 216, 219, 514 N.E.2d 1227, 1230 (1987), does hold that an employer is strictly liable for any type of sexual harassment by a supervisory employee, but it bases this holding on an Illinois statute dealing with sexual harassment, 775 ILCS 5/2-102(D), and does not mention the common law of agency. Judge Easterbrook (and Judge Wood by implication) fastens on this statute as the ground for believing that the plaintiffs in our two eases have a good claim, noting the Supreme Court’s reference in Mentor to the Restatement of Agency — though the Restatement is limited to common law, Restatement (Second) of Agency 3 (1958), so that Mentor’s reference to it is not plausibly interpret*508ed as authority for deferring to a state statute. Nor are Restatements confined to state common law, as Judge Easterbrook believes. The Mentor suit had not been filed in any state, but in the District of Columbia. The Supreme Court neither mentioned this fact nor suggested that the place to look for the agency principles applicable to the case was the law of the District of Columbia. My colleague is right that the “common law of agency has stood the test of time.” But the observation is doubly irrelevant: he forswears reliance on common law principles of agency in favor of a statute that departs from those principles; and they were not, as I shall point out, developed with reference to the novel and atypical tort of sexual harassment.
His argument implies that by a stroke of the pen Illinois could eliminate all Title VII liability in that state that depended on the principles of agency law. This would mean all Title VII liability in which the employer was not a personal participant in the infringement of the employee’s rights. A state statute that provided that a decision to fire or take other adverse action against an employee on the basis of her sex would not be imputed to the employer unless the employer’s owner or chief executive officer had approved the action would do the trick. My colleague resists this implication, arguing that an attempt by a state to curtail Title VII liability would be preempted by the federal statute. But if as he believes Congress left the formulation of agency principles to the states, there is nothing in Title VII that could conflict with a state’s decision to adopt the narrowest of such principles to govern sexual harassment or other forms of discrimination. He says we needn’t fear that states might discriminate against Title VII plaintiffs by formulating narrow agency principles because the principles would apply to claims under state law as well as under federal law. But the state statute on which he relies for the agency principles applicable to Title VII sexual harassment suits filed in Illinois is not a general agency or tort statute; it is limited to discrimination. The logic of his position is that a state hostile to discrimination claims could draft a special, and especially narrow, set of agency principles to govern discrimination cases alone.
Just as I do not think that Title VII’s silence with respect to the applicable agency principles points us toward the incorporation of state law, so I do not think that its silence can tell us anything about the structure of those principles. It thus cannot tell us whether, as I believe no one thinks (and the Supreme Court seems in Meritor to have nixed, see 477 U.S. at 72, 106 S.Ct. at 2408), the employer should be strictly liable for all acts of sexual harassment on or off the premises (in Ellerth’s case, largely off) by all employees, or whether distinctions should be made on the basis of kinds of employee (such as supervisory or nonsupervisory) or kinds of act (such as demotion versus a sexual solicitation versus forcible rape), or some combination of the two.
The Second Restatement of Agency (there is no third) was promulgated 40 years ago, before Title VII was enacted and before the concept of sexual harassment had emerged as a distinct legal concept. There is nary a hint in the text or legislative history of Title VII that Congress intended to incorporate the Restatement by reference. It would have been loopy to do so. The Restatements are intended to provide a compact statement of common law principles. The essence of the common law method of rulemaking is suppleness and flexibility, facilitating adjustment to altered circumstances, and is inconsistent with treating any statement of common law principles as a petrified text.
The provisions of the Second Restatement of Agency are designed mainly for two types of case neither of which is before us. The first is the tort committed against a stranger, as where a truck driver employed by the defendant runs down a pedestrian. The second is the contract between a stranger and an agent of the defendant. The Restatement does not completely ignore torts committed by one employee against another, see § 473, but it subjects them to the “fellow servant” rule. § 474. This nineteenth-century doctrine 'excuses the employer from liability for negligent injury of one employee by another. Farwell v. Boston & Worcester R.R., 45 Mass. (4 Met.) 49 (1842). If the absorption *509of the Second Restatement of Agency into Title VII were taken seriously, courts would have to explore the implications of the fellow servant rule for the law of sexual harassment. This they rarely do (but see Guess v. Bethlehem Steel Corp., 913 F.2d 463, 465 (7th Cir.1990)), though, as it happens, the position they have arrived at with respect to harassment by eoworkers is consistent with the rule. As I shall note in a moment, an employer is not automatically liable for coworker harassment; there is employer liability only if the employer is negligent.
The fellow servant rule might be thought inapplicable to sexual harassment because the rule is invariably stated in terms of the employee’s negligent rather than intentional torts, and sexual harassment is intentional on the part of the harasser, that is, of the fellow employee. But the only reason the rule seems to be just about torts of negligence is that the liability of an employer for his employees’ intentional torts is so much more limited than his liability for their unintentional (usually, negligent) torts that the employer would rarely have occasion to invoke the rule other than in the negligence setting. The fellow servant rule was applied to an intentional tort in Gabrielson v. Waydell, 135 N.Y. 1, 31 N.E. 969 (1892), and while it is an old case, most fellow servant cases are old, the fellow servant rale having largely gone down the drain when employers’ liability acts and workmen’s compensation acts superseded the common law of employer liability for industrial accidents. See Pomer v. Schoolman, 875 F.2d 1262 (7th Cir.1989), for a rare modern case.
If the fellow servant rale were sought to be applied to harassment by a supervisory employee, we would have to decide whether the harasser was a “vice-principal,” W. Page Keeton et al., Prosser and Keeton on the Law of Torts § 80, p. 572 (5th ed.1984), exercising a nondelegable duty of care, or merely a “superior ... servant.” Restatement, sv/pra, § 479. The employer would be liable for the vice-principal’s harassment but not the superior servant’s. The issue of classification would be made difficult to the point of unreality by the fact that the Restatement nowhere addresses sexual harassment of one employee by another, or indeed sexual harassment of any sort. It is impossible to tell from its discussion of nondelegable duties whether the supervisors in our two cases would have been thought vice-principals.
Sexual harassment differs from most other torts, and especially from the most common torts — the deliberate or accidental infliction of physical injury — in being “invisible” or largely so other than to victim and perpetrator. The employer will know whether one of his employees sustains a physical injury on the job; he is much less likely to know that one of his employees is being sexually harassed on the job. The feasibility of an employer’s controlling the torts of his employees is, as I shall show, an important consideration in crafting agency principles, and one dependent on the particular tort.
I may be in a minority of judges in not believing (or at least in being willing publicly to deny) that the most fruitful way of approaching a novel issue of law is through seeking analogies in legal doctrines designed with quite other issues in mind. Nevertheless, had the Supreme Court told us to use the Second Restatement of Agency as the framework for evaluating sexual harassment under Title VII, I would bow to its command. It did not; but by citing the Restatement it gave lawyers and judges a straw to grasp at. The straw has broken in their hand. The Restatement turns out to be hopelessly vague in its bearing on the issue of employers’ liability for sexual harassment, being vaguely worded and addressed to other issues. So judges can in good faith reach opposite results when they seek guidance in the Restatement to. employers’ liability for sexual harassment by supervisory employees. See, e.g., Faragher v. City of Boca Raton, 111 F.3d 1530 (11th Cir.1997) (en banc). The judges and lawyers who insist that the Restatement of Agency is The Way either are disingenuous, wishing to conceal their true grounds of decision, or are in the grip of the formalist belief that difficult cases can be decided by teasing out the meaning of words in a text composed with other problems in mind, without need to examine the social policies that the law might be thought to be serving. “We condemn Lochner as formalis*510tic not because it involves a choice,'but because it attempts to describe this choice as compulsion.” Frederick Schauer, “Formalism,” 97 Yale L.J. 509, 511-12 (1988). Since neither the text nor the legislative history of Title VII indicates what agency principles the authors of the statute had in mind, the formalist gropes for another text, finds the Restatement, and treats it inappropriately as a surrogate statute.
It is time that we threw away the crutch of the Restatement and, recognizing the differences between workplace sexual harassment and the actual subject matter of that antiquated screed, ask as an original matter — as I think the Supreme Court invited us to do in Meritor when it (under)stated that the common law of agency might not be fully transferable to sexual harassment — what the best regime of liability would be for these cases. It would be misleading even to speak of applying to this case “the law of agency.” That would imply the existence of some ready-at-hand body of rules that we have only to apply to a new set of circumstances. The problem is not application; it is creation. I think it equally question-begging to ask whether a supervisor’s conduct is “attributable” to the employer, or whether he may be said to “speak for” the employer, or whether an employer must be deemed to “delegate his common law duties” to his supervisory employees. These are just additional ways of asking whether the employer should be liable.
I understand the attraction of seeking a textual answer elsewhere, perhaps in local law, as Judge Easterbrook suggests; but this evasion of the occasional duty of judicial creativity will not work either, as I have explained and Judge Manion more fully. We thus cannot avoid the task of trying to create a set of agency principles that will deter sexual harassment without imposing an unreasonable burden on employers. I do not consider the burden of liability on employers a negligible consideration in formulating federal common law rules to govern aspects of employment regulation. Our labor markets are becoming choked by regulation, all well meaning but cumulatively an impediment to the efficient employment of the nation’s most valuable economic resource, which is its workers.
I emphasize deterring sexual harassment rather than compensating its victims because, unlike many torts — and again the clearest illustrations are torts that inflict physical injury — sexual harassment does not usually bring about a significant change in the victim’s wealth. The victim may be humiliated and deeply distressed by it; rarely will she (or the very occasional he) be impoverished by it. This is not to say that psychological pain should not figure in the calculation of damages. It should. All I wish to emphasize is that an award of damages in these cases is not primarily designed to protect the victim’s standard of living. The payment of damages in the usual case of sexual harassment is an instrument for deterring future incidents of such harassment rather than for restoring lost earnings or for financing expensive curative or rehabilitative measures.
There is no novelty in grounding principles of agency on deterrence. A recognized objective of imposing “vicarious” liability for the torts of employees (the liability that goes by the name of “respondeat superior”) is to reduce the amount of tortious behavior. Hartmann v. Prudential Ins. Co., 9 F.3d 1207, 1210 (7th Cir.1993); Konradi v. United States, 919 F.2d 1207, 1210 (7th Cir.1990); Alan O. Sykes, “The Boundaries of Vicarious Liability,” 101 Haro. L. Rev. 563, 570 (1988). Our court has adopted (and it is not alone in doing so) the deterrent perspective as its guide to dealing with cases involving the creation of a hostile environment by the plaintiffs coworkers. The standard of employer liability in those cases is negligence: an employer who knew or should have known about the harassment and failed to take reasonable remedial measures is liable. McKenzie v. Illinois Dep’t of Transportation, 92 F.3d 473, 480 (7th Cir.1996); Baskerville v. Culligan International Co., 50 F.3d 428, 432 (7th Cir.1995); Carr v. Allison Gas Turbine Division, 32 F.3d 1007, 1009 (7th Cir.1994); Doe v. Lago Vista Independent School District, 106 F.3d 1223 (5th Cir.1997); Rosa H. v. San Elizario Independent School District, 106 F.3d 648, 656 (5th Cir.1997); *511Splunge v. Shoney’s, Inc., 97 F.3d 488, 490 (11th Cir.1996); Burns v. McGregor Electronic Industries, Inc., 989 F.2d 959, 966 (8th Cir.1993). The thinking is that if the employer has a realistic opportunity to prevent harassment, he must take it.
But if the goal is deterrence, why, it may be asked, is the standard negligence rather than strict liability? The reason is the infeasibility of an employer’s stamping out this sort of harassment without going to extreme expense and greatly curtailing the privacy of its employees, as by putting them under continuous video surveillance. If the victim of sexual harassment complains to a supervisor, or a worker who notices what’s going on complains to a supervisor, or the harassment is so pervasive (considering its nature, frequency, and number of victims and perpetrators) that the employer knows or should know about the harassment, the employer ought to take steps to correct the problem. He has or should have the information; he has only to act upon it. And since everyone knows by now that sexual harassment is a common problem in the American workplace, the employer ought in addition to take, in advance of specific cases of harassment, preventive measures against it, as by adopting and announcing a policy against sexual harassment and creating a discreet and convenient machinery by means of which victims can obtain relief without exposing themselves to retaliation. These are the responsibilities that a negligence standard imposes. They should deter most sexual harassment by coworkers without the courts’ subjecting the employer to strict liability. A law that requires the employer to do more than is feasible to control harassment will impose costs without creating deterrent benefits. In the long run, these costs will be borne largely by consumers, in the form of higher prices for the employer’s product, and workers, in the form of lower wages (because the higher costs are labor costs). Many consumers and workers are women, so women, who are the principal victims of sexual harassment, will pay a big part of the costs that employers incur as a consequence of excessively harsh principles of employers’ liability.
In these circumstances, strict liability would not only be expensive and unnecessary, and possibly regressive as well; it would be futile. If the law imposes liability in harassment cases in which there is no reasonable measure that the employer could have taken to prevent the harassment, the only effect of the law will be to impose extra costs on employers and those with whom they are linked contractually. Employers will prefer paying the occasional judgment to incurring costs that, by definition, exceed the employer’s foreseeable liability — by definition because, were the costs less than the expected liability, the failure to incur them would be negligence; it is only when they are greater, so that the employer would not be negligent for failing to incur them, that strict liability bites. The bite may come out of the hides of innocent workers and consumers, rather than deterring any more harassment than a negligence standard would do.
What this shows is that strict liability doesn’t always have the happy result of just deterring a bad practice more effectively than the negligence standai’d. Sometimes the incremental contribution of strict liability to deterrence is nil or slight, and easily outweighed by the additional costs of a more expansive liability. This is implicit in the rejection of strict liability for coworker harassment cases, a rejection to which I believe all members of this court subscribe.
But the cases in which the issue is the employer’s liability for harassment by coworkers are the easy ones. The difficult ones are those in which, as in the cases before us, the harasser is a supervisor rather than a line employee. We must distinguish between two types of sexual harassment by supervisors. In the first type, the supervisor uses or attempts to use his supervisory authority to obtain sexual favors from an employee. This is the domain of what has come to be called “quid pro quo” harassment. In the second type of case, the supervisor does not use or attempt to use his supervisory authority at all. He harasses an employee, but he does so in exactly the same way that an employee who had no supervisory authority would harass another employee. He uses unwanted terms of endearment; he fondles *512or rubs up against his victim; he displays sex toys or tells dirty jokes; he brags about his sexual skills; he proposes marriage; he threatens to kill himself; in the extreme case, illustrated by Martin v. Cavalier Hotel Corp., 48 F.3d 1343 (4th Cir.1995), he rapes her. These are modes of harassment equally available to a nonsupervisory employee. The proper standard of employer liability here is negligence, just as in the ease of harassment by nonsupervisory employees, because it is as costly for the employer to police this kind of harassment by a supervisor as it is to police the identical harassment by a coworker. I believe that we all have qualms about Martin, the case that carries strict liability for the acts of a supervisory employee to its logical extreme.
Although the standard should be the same in the two types of hostile-environment case, the application of the standard is bound to differ as a practical matter. Even if the supervisor is scrupulous in forbearing to invoke his authority over the employee he is harassing, the existence of that authority is a fact of which the victim is bound to be constantly aware. She is more likely than in a case of coworker harassment to submit and less likely to complain. She may believe that because her harasser is a supervisor he is highly valued by the company and that as a consequence the higher-ups in the company will reject her complaint — will in fact fire her, not the harasser. Because such fears will often be reasonable, due care on the part of employers to prevent sexual harassment by its supervisory employees may require the creation of additional complaint machinery when the complaint is against a supervisor. Whether the employer had complied with this heightened duty of due care would be a jury issue in the first instance.
But I would not go further and impose strict liability on the employer for hostile-environment harassment by a supervisor. The “theory” that by making him a supervisor the employer had clothed him with the appearance of authority to harass employees is empty. Since everyone knows by now that sexual harassment is unlawful, a company that publicizes and enforces a stern policy against sexual harassment has done all that it can reasonably be expected to do to dispel its supervisors’ “apparent authority” (to lapse into Restatement-speak) to harass. The adequacy of the policy in the circumstances (what if the harasser is the chief executive officer of the employer?) is a factor to be considered in making the required determination of negligence. There is no need to have a separate standard just because the harasser is a supervisor. Excessive complexity is the bane of American law; we have an opportunity to make it a little simpler.
So much for hostile-environment harassment. Let me turn now to quid pro quo harassment. We have to distinguish between two types of such harassment. In the first, the supervisor brings about a significant alteration in the terms or conditions of his victim’s employment. He fires her, or denies her a promotion, or blocks a scheduled raise, or demotes her, or transfers her to a less desirable job location, or refuses to give her the training that the company’s rules entitle her to receive. In all these examples the supervisor is using his delegated authority to do a company act. Strict liability is appropriate in this kind of case, because it is likely to deter this kind of sexual harassment more effectively than negligence liability would. The employer who is strictly liable will monitor the exercise of this delegated authority very carefully, knowing that it will be liable if the authority is abused. And this monitoring should be relatively easy to do. For it is usually a mistake for a firm, quite without regard to any potential legal liability, to give a supervisor unilateral authority to alter a subordinate’s terms or conditions of employment significantly. In well-managed companies, decisions having such consequences are subject to rules, and to review by higher-ups in the company — the industrial equivalent of appellate review. The rules will be more carefully formulated and the supervisor’s compliance with them in firing or otherwise hurting a subordinate more carefully reviewed by the supervisor’s superiors if the employer is strictly liable for the supervisor’s use of his delegated powers to harass subordinates. Courts applying a negligence standard would have great difficulty determining how closely the supervisor should be supervised. Such questions as *513how many tiers of review should be provided before an employee can be fired or demoted are not easily answered in terms of reasonableness or due care, the criteria of negligence. The regime of strict liability shifts the responsibility for deciding these questions to the employer, who knows more than judges do about how to control supervisory employees.
Could the argument that I have just sketched be made in support of strict liability for all sexual harassment by company employees, including harassment by coworkers? Employers know more than we do about how to extirpate that form of workplace sexual harassment too. True; but courts know, more or less, what is reasonable for the employer to do about hostile-environment harassment — institute a tough policy, disseminate it, establish a procedure by which a worker can complain without fear of retaliation (so don’t require that the complaint be made to the man who is harassing her!), respond promptly and effectively to any report of possible harassment. Knowing what the employer should do, the courts have only to decide whether he did it, in order to decide whether he was negligent and should be liable. But when it comes to designing the optimum system for reining in the discretion of supervisory employees, the courts are at sea and it makes sense therefore to shift the responsibility entirely to the employer to create and administer an effective system for the review and control of company actions taken by supervisors in the exercise of their delegated authority.
Strict liability is inappropriate, however, when the supervisor merely makes threats, even if the threats are effective. This is why it is important to distinguish between the type of quid pro quo harassment in which the supervisor actually alters the terms or conditions of his victim’s employment and the type of harassment in which he merely threatens to do so, whether or not the victim yields to the threats. Suppose the supervisor threatens to fire a subordinate unless she’ll have sex with him, and she agrees — or refuses and he does not carry out his threat. In either case, because he has not used his delegated authority to commit a company act, there is no way in which a system for vetting such acts would catch him out. In a well-regulated company a supervisor who wants to fire a subordinate has to obtain the approval of higher-ups, as I have said, and they will have an opportunity therefore to determine the bona tides of his proposal. But if he doesn’t propose to fire her, whether because she has submitted to his sexual extortion or called his bluff, there will be no proposed action to review. It will be no more feasible for the company to determine what is going on than it would be if the harasser were a coworker who had threatened to steal the victim’s work tools if she didn’t submit to him.
Romantic encounters, including romantic encounters between supervisors and supervised, are a fact of the workplace. Title VII does not purport to forbid them, and would be quixotic if it did. Many happy marriages have grown out of such encounters. Some of the encounters are abusive from the start, and some start well and turn ugly and engender charges of sexual harassment that sometimes have and sometimes lack merit. The words, the gestures, the other behaviors that differentiate the fully consensual relationship from the coercive relationship will often be invisible to the supervisor’s superiors. The yielding to a threat will look no different from the yielding to a lawful proposal. It is only when the threat is carried out that the abusive supervisor does something, such as firing the supervised employee, that the employer will know about and should monitor. It is facile to suggest that employers are quite capable of monitoring a supervisor's actions affecting the work environment. Large companies have thousands of supervisory employees. Are they all to be put under video surveillance? Subjected to periodic lie-detector tests? Trailed on business trips by company spies? Even the EEOC does not go so far. The culminating recommendation in its “Guidelines on Discrimination Because of Sex” to employers is that they should “deveIop[ ] methods to sensitize all concerned” to the problem of sexual harassment. 29 C.F.R. § 1604.11(f). This is easy to say, but, as the armed forces have recently discovered to their sorrow, hard to do. It is unrealistic to expect that adopting strict liability in lieu of negligence, a relative*514ly esoteric and marginal change in the law, will do much to create a corporate culture in which sexual harassment is unthinkable. Granted, the problem of monitoring the non-company acts of supervisory employees is primarily a problem of large employers, for in companies that are at or near the statutory minimum of 15 employees, 42 U.S.C. § 2000e(b), noncompany acts will presumably be visible to top management, creating de facto strict liability. But it is the large companies that are the usual targets of litigation, because of their deep pockets.
The distinction that I am proposing between the harassing supervisor who commits a company act, such as firing, and the supervisor who merely threatens such acts, has at least a footing in the case law. See Gary v. Long, 59 F.3d 1391, 1396 (D.C.Cir.1995); Sauers v. Salt Lake County, 1 F.3d 1122, 1127 (10th Cir.1993); Kotcher v. Rosa & Sullivan Appliance Center, Inc., 957 F.2d 59, 62 (2d Cir.1992); Hicks v. Gates Rubber Co., 833 F.2d 1406, 1414 (10th Cir.1987). There are hints of it in Judge Easterbrook’s and Judge Kanne’s separate opinions. It is, for sure, not the orthodox approach, see, e.g., Karibian v. Columbia University, 14 F.3d 773, 778 (2d Cir.1994); Highlander v. K.F.C. Nat’l Management Co., 805 F.2d 644, 648 (6th Cir.1986); I would go so far as to concede that it has the curse of novelty; but the cases that take the orthodox approach do not explain why they do so, and the approach that I am proposing in its stead is not foreclosed by any decisions of the Supreme Court. The Court in Meritor said that Congress intended “to place some limits on the acts of employees for which employers under Title VII are to be held responsible.” 477 U.S. at 72, 106 S.Ct at 2408. It didn’t tell us what those limits are.
We can test my proposed approach by considering its application to other forms of discrimination forbidden by Title VII. The closest parallel is sex discrimination that does not involve harassment in the usual sense. Compare two cases. In one, a supervisor denies a raise to a subordinate because of her sex. In another, the supervisor slaps her because of her sex. In the first case, the supervisor has done a company act, and the company is strictly liable. In the second case, the supervisor has not done a company act, and there is no liability under Title VII unless the employer was negligent in hiring or failing to fire or supervise the employee. In Restatement-speak, the supervisor in the second case has not acted in the employer’s interest; but that is just a conclusion. The practical reason why the employer is not liable is that the supervisor’s act took place outside the channels in which action by supervisors is realistically controllable by their superiors.
The distinction between company and noncompany acts points the way to the proper handling of a theoretical third category of supervisor sexual harassment. That is where the supervisor uses his supervisory authority to create a hostile work environment, perhaps by encouraging his male subordinates to harass his female subordinates. If he does this without committing any company acts, then the case should be assimilated to the ordinary case of noncompany harassing acts by a supervisor, and the employer’s liability should be governed by negligence. If the supervisor commits company acts, then the employer should be strictly liable. As a practical matter, the employer will be hable in almost any case in which a supervisor is orchestrating a campaign of sexual harassment, for the employer will either learn of the campaign or be negligent in failing to notice it.
I do not think it is an objection that my proposed approach might occasionally require applying different standards of employer liability to different conduct of the same supervisory employee, whose sexual harassment of a subordinate might include both company acts and noncompany acts. Suppose a worker got drunk at a company party and on his way out of the building bumped into and injured a customer, then got into his car and on his way home hit a pedestrian, both accidents being due to his drinking. The company would be strictly liable to the customer — respondeat superior would apply — but the pedestrian would have to prove the company negligent in order to establish its liability to him, because the drive was not within the scope of his employment.
*515My proposed approach is summarized in the following table. The term “company act” signifies an act that significantly alters the terms or conditions of employment of the victim of sexual harassment and “noncompa-ny act” signifies hostile-environment harassment by coworkers or supervisors or the kind of quid pro quo harassment that involves only unfulfilled threats (either because the victim submits or because she calls the supervisor’s bluff), so that no company act is committed. The difficult borderline case is that of constructive termination precipitated by a threat. The termination will look to the supervisor’s superiors like a voluntary quit. But since there is always some paperwork involved in an employee’s quitting, the higher-ups in the company will have some ability to monitor constructive discharges, and I would therefore impose strict liability in such cases.
Liability for Sexual Harassment Under Title VII: A Suggested Approach “Company act” “Noncompany act” By supervisor employer strictly ' liable employer liable only if negligent By coworker N.A. employer liable only if negligent
I now apply the suggested approach to the facts of our two cases. In Jansen there was no company act and therefore no quid pro quo violation. Antoni did hold up Jansen’s performance review and raise for several months, but the review was favorable, the raise was made retroactive, and Jansen herself admits that delays of the length she encountered were commonplace. The delays she encountered were not significant alterations in the terms or conditions of employment — therefore were not “company acts” as I am using that term in order to denote the proper scope of quid pro quo liability. For they were not visible to and therefore corri-gible by higher-ups in the company. The tenuous character and trivial consequences of Antoni’s threat furnish a practical argument for the limited scope of quid pro quo liability that I am urging.
Jansen has, however, presented enough evidence to establish a prima facie case of hostile-environment harassment. What Antoni did to Jansen over a period of some eighteen months — a barrage of threats and sexual solicitations — was highly offensive, and not mere sexual banter or even the kind of mildly obnoxious vulgarity that we held in Baskerville v. Culligan International Co., supra, did not rise to the level of actionable harassment. The only issue is whether Packaging Corporation should have discovered what was going on. The answer is yes, a tentative yes but one that can be given with enough confidence to defeat summary judgment for the employer. Two previous complaints of sexual harassment had been lodged against Antoni. The second, it is true, was somewhat cryptic; but the company’s response to it, which was to counsel Antoni about the company’s sexual harassment policy, showed that Packaging Corporation “read” the complaint as one of sexual harassment, as it pretty obviously was. Antoni was also well known to be having sex with another worker, and while this relationship was apparently consensual, the fact that he was married to someone else at the time should have alerted the company to the possibility that he was sexually incontinent. The company’s lackadaisical if not perverse reaction to Jansen’s complaint (accusing her of sexual harassment) is further evidence of the company’s negligence with regard to the harassment of female employees.
So the grant of summary judgment to Packaging Corporation with regard to the charge of sexual harassment should be reversed. But the only issue for determination by the jury on remand should be whether the company was negligent in failing to protect Jansen from Antoni. There is, moreover, a serious question whether she can obtain substantial damages, given her failure to complain for eighteen months. We have held that the doctrine of avoidable consequences, on which see, e.g., Outboard Marine Corp. v. Babcock Industries, Inc., 106 F.3d 182, 184 (7th Cir.1997); Ellerman Lines, Ltd. v. S.S. President Harding, 288 F.2d 288, 289-90 (2d Cir.1961) (Friendly, J.); Keeton et al, supra, § 65, pp. 458-59, is applicable to violations of Title VII as to other torts. Hunter v. Allis-Chalmers Corp., 797 F.2d 1417, 1427 (7th Cir.1986). The precise issue is whether if Jansen had complained sooner the company would have taken effective action. This is *516uncertain given the lackluster way in which the company responded when she did complain. The uncertainty creates an issue for a jury.
On my view of the scope of quid pro quo liability, Ellerth like Jansen is purely a hostile-environment case. Slowik was Ellerth’s supervisor. It is of no consequence, so far as his supervisory authority over her was concerned, that there was another supervisor in between them; this just shows how much power he had over her, given that he was that much higher in the corporate hierarchy. But he never committed a company act against her, such as denying her a promotion, although he obliquely threatened to do so. Therefore she should have to show that the company was negligent in failing to detect and prevent the harassment. The evidence of negligence is too thin to create a jury issue — so it is no surprise that Ellerth doesn’t even argue that Burlington was negligent. Although aware of the company’s policy against sexual harassment, Ellerth did not complain about Slowik to anyone in the company’s human resources department or to anyone else who might have taken corrective action; she did not complain to her supervisor. Unlike Jansen’s employer, Burlington Industries had no other sources of information about its supervisor’s misbehavior. The fact that Slowik told off-color jokes and commented on Ellerth’s legs at a lunch in which a vice-president of Burlington was present could not be thought to put Burlington on notice of probable harassment. The vice-president had no reason to think Ellerth offended by Slowik’s banter; not all women would be. (Slowik had also patted Ellerth’s knee under the table at that lunch, but there is no indication that this was visible to the vice-president.) Although there is a bit of evidence that Slowik may have harassed another woman employed by Burlington, there is no evidence that this other harassment came or should have come to the attention of any higher-ups in the company. The fact that Ellerth complained to a few of her fellow non supervisory employees (and toward the end to a customer service manager) also did not put the company on notice. The precedents in our court and the other circuits do not authorize the basing of a finding of negligence on such scraps of evidence. See, e.g., Doe v. R.R. Donnelley & Sons Co., 42 F.3d 439, 447 (7th Cir.1994); Juarez v. Ameritech Mobile Communications, Inc., 957 F.2d 317, 320-21 (7th Cir.1992); Murray v. New York University College of Dentistry, 57 F.3d 243, 250 (2d Cir.1995). A case in which the plaintiff does not think enough of her ability to prove negligence to even raise a negligence claim would be a pretty dubious vehicle for going beyond the precedents.
But it is an even weaker case for negligence than I have yet let on. Ellerth’s claim that she feared that complaining about Slowik would cost her her job is inconsistent with her claim that she quit the company because Slowik had made her working conditions unbearable. If the choice is to quit the company or complain, you obviously should complain, since if the complaint gets you in trouble you’re no worse off than you would have been quitting — unless you are seeking a reference for your next job, and there is no evidence of that in Ellerth’s case.
Her action in quitting rather than complaining underscores the importance of agency principles that place appropriate pressure on victims of sexual harassment to protect themselves from harassment by complaining to the harasser’s superiors. Tort law in general allocates responsibility for the avoidance of nonintentional injuries (not just for their consequences) to both injurer and victim, rather than to the injurer alone. Sexual harassment is an intentional injury by the harasser, but an unintentional injury (in the usual case and in the two cases before us) by the harasser’s superiors, or in other words by the employer itself. The employer is innocent; the victim is innocent; the optimal system of liability for minimizing sexual harassment requires cooperation by both. Once Ellerth decided that Slowik had made the workplace unendurable, she had nothing to lose and everything to gain by complaining. It follows that as between her and her employer, she was in the better position to prevent the misconduct of her supervisor. Slowik had a responsibility to his principal not to harass Ellerth, but Ellerth had a responsibility to her principal too — a duty to complain about Slowik through proper chan-*517neis. With rights should come responsibilities.
In any event, Ellerth explicitly waived negligence as a ground for Burlington’s liability, and so it is not available as a basis for reversing the judgment in Burlington’s favor. In her reply to the petition for rehearing, repeating what she had said earlier in her brief on appeal, Ellerth states, “Appellant [Ellerth], in fact, agrees with Burlington that § 219(2)(b) does not apply to this case” (emphases in original). The reference is to section 219(2)(b) of the Restatement — the section that makes an employer liable for the torts of his employee if the employer is negligent.