Robert Johnson hanged himself in his cell in the Metropolitan Correctional Center, the federal jail in Chicago, where he had been held for six months awaiting trial for extortion. His estate brought this suit for damages against the United States under the Federal Torts Claim Act, 28 U.S.C. §§ 1346(b), 2671-80, charging that the jail was negligent in failing to identify him as a suicide risk and take suitable preventive measures. The Act incorporates local law, id., § 1346(b), and so the question is the liability of the United States under the principles of Illinois tort law, though so far as bears on issues of liability in this case they are the general principles of Anglo American tort law, not anything special to Illinois.
The district judge, after a bench trial, awarded the estate $1.8 million in damages, of which $1.6 million was intended to compensate for the pain and suffering that Johnson suffered during the period of several minutes in which he remained alive before strangling while hanging from the noose that he had made out of a bedsheet and suspended from an exposed pipe that ran beneath the ceiling of his cell. The other $200,000 in damages were for the loss to his family resulting from his death.
Although the initial psychological screening that Johnson like all new inmates underwent revealed no history of psychiatric disorders or suicide attempts and no suicidal thoughts, and although this was not his first spell of incarceration, he deteriorated markedly during his stay in the jail. He scratched and picked at sores *755on his body until they bled so copiously that they stained his clothes and bedsheets and were noticed by other inmates. One of the guards observed that Johnson was a loner, slept a lot during the day, seemed nervous, had poor hygiene, and had a lot of sores that bled; yet the guard did not think to refer Johnson to the medical or psychology department of the jail. Another inmate told at least one other member of the prison staff that Johnson had a nervous problem and might need an antidepressant, but nothing was done about this information either. The day before Johnson killed himself his cellmate importuned him to see a physician’s assistant and filled out a sick-call form in which the cellmate stated: “Open sores all over body caused from nerves. Request to see a psychologist or psychiatrist.” Johnson presented the form to the physician’s assistant the next day, but she neglected to read it and when he told her his roommate was taking some kind of medication that he thought he should be taking as well she merely told him to make an appointment to see the jail psychologist. He did not do that, but instead hanged himself 12 hours later.
The government does not deny that the jail’s staff was negligent in failing to discover that Johnson had some kind of nervous condition and to take steps to deal with it. The obsessive scratching and picking, in the context of a general pattern of abnormal behavior, should have alerted the staff to the fact that Johnson might have a psychiatric illness. The physician’s assistant whom Johnson consulted should have read the sick-call form and had she done so she might well have sent him directly to the jail psychologist rather than relying on him to make an appointment to see the psychologist. But it was the plaintiffs burden to prove both that Johnson would not have committed suicide had the jail’s staff acted responsibly, e.g., Beul v. ASSE Int’l, Inc., 233 F.3d 441, 445-47 (7th Cir.2000); Merco Distributing Corp. v.Commercial Police Alarm Co., 84 Wis.2d 455, 267 N.W.2d 652 (Wis.1978); Guthrie v. American Protection Industries, 160 Cal.App.3d 951, 206 Cal.Rptr. 834, 836 (Cal.App.1984); Vastola v. Connecticut Protective System, Inc., 133 Conn. 18, 47 A.2d 844, 845 (Conn.1946), and that his suicide was a foreseeable as well as actual consequence of the staffs negligence.
The issue of causation is doctrinally straightforward; not so the issue of the foreseeability of suicide. When failure to prevent a suicide is claimed to be negligent, the issue of foreseeability is analyzed under the rubric of “supervening cause” and the general rule is that the negligent actor is not liable for the victim’s decision to kill himself. The suicide is said to be a supervening cause of the victim’s loss of his life, breaking the chain of responsibility that would otherwise link the loss to the negligent act. E.g., Beul v. ASSE Int’l, Inc., supra, 233 F.3d at 447; McMahon v. St. Croix Falls School District, 228 Wis.2d 215, 596 N.W.2d 875, 879 (Wis.App.1999); Wyke v. Polk County School Board, 129 F.3d 560, 574-75 (11th Cir.1997); Bruzga v. PMR Architects, P.C., 141 N.H. 756, 693 A.2d 401 (N.H.1997); Edwards v. Tardif, 240 Conn. 610, 692 A.2d 1266, 1269 (Conn. 1997); W. Page Keeton et al., Prosser and Keeton on the Law of Torts § 44, p. 311 (5th ed.1984). Of course this is just a conclusion, not reasoning; but it is a conclusion sustained by reasoning about the unforeseeability of most suicides and the role of foreseeability in determining tort liability. If an employer refuses an employee’s request for a raise, the latter may respond by killing himself, and yet the employer even if somehow negligent in failing to givé the employee the raise would not be legally responsible for the death, just as if through the carelessness of the driver a truck spilled a toxic sub*756stance and a passerby scraped it up ánd poisoned his mother-in-law with it the driver would not be liable to the mother-in-law’s estate; the son-in-law’s criminal act would be deemed a supervening cause. See Rowe v. State Bank of Lombard, 125 Ill.2d 203, 126 Ill.Dec. 519, 531 N.E.2d 1358, 1361 (Ill.1988); Giebel v. Richards, 224 Wis.2d 468, 591 N.W.2d 901 (Wis.App. 1999); Henry v. Merck & Co., 877 F.2d 1489, 1494-97 (10th Cir.1989); Shelton v. Board of Regents, 211 Neb. 820, 320 N.W.2d 748, 752-53 (Neb.1982).
A person is not liable for such improbable consequences of negligent activity as could hardly figure in his deciding how careful he should be. Liability in such circumstances would serve no deterrent, no regulatory purpose; it would not alter behavior and increase safety. Nothing would be gained by imposing liability in such a case but compensation, and compensation can be obtained more cheaply by insurance. But by the same token the doctrine of supervening cause is not applicable when the duty of care claimed to have been violated is precisely a duty to protect against ordinarily unforeseeable conduct. A risk unforeseeable to an ordinary person is foreseeable to a specialist who assumes a duty to prevent the risk from materializing. The duty is a recognition that the unforeseeable has become foreseeable to the relevant community. And so a hospital that fails to maintain a careful watch over patients known to be suicidal is not excused by the doctrine of supervening cause from liability for a suicide, e.g., DeMontiney v. Desert Manor Convalescent Center, 144 Ariz. 6, 695 P.2d 255, 259-60 (Ariz.1985), any more than a zoo can escape liability for allowing a tiger to escape and maul people on the ground that the tiger is the supervening cause of the mauling. City of Mangum v. Brownlee, 181 Okla. 515, 75 P.2d 174 (Okla.1938); see also Scorza v. Martinez, 683 So.2d 1115, 1117 (Fla.App.1996); Behrens v. Bertram Mills Circus, Ltd., [1957] 2 Q.B. 1, 1 All E.R. 583 (1957). In both cases there is a foreseeable, in ■ the sense of probable, hazard which precautions can and should be taken in order to lessen. So we may set the language of supervening cause to one side and ask simply whether Johnson’s suicide was a foreseeable consequence of the negligence of his jailers in responding ineffectually to his abnormal behavior.
But first we should consider whether there was any causal relation between that negligence and the suicide. We think not. Johnson did not commit suicide because he had sores on his body. As nearly as can be reconstructed from the evidence, he committed suicide because he was upset about being in jail, separated from his family (including a daughter born while he was in jail) and facing a prison term if he was convicted. The scratching and the suicide were the consequences of an underlying unhappiness. Nothing the jail’s staff could have done would have alleviated Johnson’s concerns about separation and imprisonment. They were inherent in his situation. True, had the jail’s psychologist interviewed him, Johnson might have expressed suicidal thoughts and in that event the jail would doubtless have placed him on suicide watch and thereby prevented him from killing himself. (At least it would have moved him out of a cell that, because of the exposed pipe near the ceiling, made it easy for an inmate to hang himself.) “Might” is the operative word. Johnson had no history of psychiatric illness and had not revealed any suicidal ideation at his intake exam. Until nudged by his roommate to visit the physician’s assistant he had made no effort to contact the medical (including psychiatric) personnel of the jail. It is sheer conjecture that an interview with the jail psychologist would have produced sufficient information *757to have enabled tbe psychologist to infer that Johnson was a suicide risk and place him on suicide watch. The psychologist would have noticed Johnson’s sores— would even, we may assume, have pronounced him or his behavior bizarre. But bizarre behavior and suicidal behavior are different, and there is no evidence that suicidal tendencies can be inferred from the kind of behavior that Johnson exhibited.
It is true that jail inmates are much more likely to commit suicide than free persons are — in fact, nine times as likely. Lindsay M. Hayes & Joseph R. Rowan, National Sttidy of Jail Suicides: Seven Years Later 54 (National Center on Institutions and Alternatives, Feb. 1988). According to the study just cited (the only one we have found), 12.9 percent of jail suicides occur within the first three hours of confinement, 32.8 percent within the first 24 hours, 62.1 percent within the first two weeks, 72.8 percent within the first month, 89.2 within the first four months, and 97.4 within the first seven months. The likelihood of suicide is highest between 2 and 14 days of confinement, and only 10.8 percent of suicides occur after 5 months. Id. at 36. Thus, when Johnson was in his worst psychological state,' after 5 months of being in jail, the statistical likelihood of suicide was diminishing rapidly. No evidence has been offered that a prisoner who has no history of suicide attempts or even suicidal thoughts should be considered a suicide risk merely because, months after his imprisonment began, he displays symptoms of a possible psychiatric disturbance.
If the jail psychologist would not have diagnosed Johnson as suicidal, there is no reason to believe that his suicide would have been averted. It is here that the analysis of causality and the analysis of foreseeability, though distinct (the issue of foreseeability does not arise until the defendant’s negligent act is identified as a cause of the plaintiffs injury), blend insensibly into each other. The reason that measures to prevent Johnson’s suicide probably would not have been taken even if the jail’s staff had been exercising due care is that the suicide was not a foreseeable consequence of Johnson’s behavior. The population of prisons and jails is not a random sample of American citizens. It is largely a subset of the criminal population (not entirely, since some pretrial detainees are innocent of the crimes for which they are awaiting trial), itself a population prone to abnormal behavior, and the conditions of incarceration place the prisoners under considerable psychological strain. Abnormal behavior in jails and prisons is therefore-common. Whether such behavior connotes a substantially enhanced suicide risk after the first few days or weeks or months of incarceration is a question on which the record of this case, including the expert evidence, casts no light. Had Johnson developed infections from his obsessive picking and scratching, the government might well be liable; for infection is a foreseeable result of that behavior. Suicide, so far as the record shows, is not.
The psychiatric condition most closely associated with suicide is severe (“clinical”) depression, and a disturbed sleep pattern is a symptom of depression. But of course many people who sleep badly are not depressed, and most people who are depressed do not commit suicide. No evidence was presented that compulsive scratching is a symptom of depression. Had the staff been more alert, maybe they would have stumbled on some underlying psychiatric condition predictive of suicide. But maybe not, because no one knows whether Johnson committed suicide because he was depressed or because he could not stand the prospect of imprisonment — as the district judge well understood.. He asked “why did Bob Johnson take his own life” and answered that “the *758evidence presented'some reasons but does not establish any of them with certainty. While an official declaration by this Court might be useful in trying to come to grips with this human tragedy, no explanation will be offered because it would necessarily be dependent upon an inappropriate degree of speculation.” Precisely. The judge should have ended his opinion there. Since no one knows why Johnson killed himself, there is no basis for thinking that due care by the staff would have prevented him from doing so (by detecting the risk and, as a consequence, placing him on suicide watch), or that, the staff should have foreseen that suicide would be a likely result of its being careless.
Although our analysis makes it unnecessary to consider the government’s objection to the size of the award of damages for pain and suffering, we shall address that objection in an effort to provide some guidance for future cases. The judge awarded the plaintiff $1.6 million for Johnson’s pain and suffering without any explanation of how he had arrived at this amount, apart from noting that “there is no dispute that Bob Johnson did not die instantly.” Juries do not explain their reasoning process, but Fed.R.Civ.P. 52(a) requires judges, when they are the triers of fact, to make written findings in support of their conclusions. This means, when the issue is the amount of damages, that the judge must indicate the reasoning process that connects the evidence to the conclusion, here the évidence that Johnson may have remained conscious for anywhere from three to twenty minutes while hanging and the conclusion that his estate should receive $1.6 million for his pain and suffering. The judge should have determined what the best estimate of the amount of time that Johnson remained conscious was and explained why (supposing it was four minutes) an award of $400,000 for each minute of pain and suffering is appropriate. He did neither.
Awarding any amount of damages for pain and suffering has long been criticized as requiring the trier of fact to monetize a loss that is incommensurable with any monetary measure. We do not agree with the criticism. Pain and suffering are perceived as costs, in the sense of adversities that one would pay to be spared, by the people who experience them. Unless tort-feasors are made to bear these costs, the cost of being adjudged careless will fall and so there will be more accidents and therefore more pain and suffering. The problem of figuring out how to value' pain and suffering is acute, however. Various solutions, none wholly satisfactory, have been suggested, such as asking the trier of fact, whether jurors or judge, to imagine how much they would pay to avoid the kind of pain and suffering that the victim of the defendant’s negligence experienced or how much they would demand to experience it willingly, 2 Dan B. Dobbs, Dobbs Law of Remedies: Damages Equity Restitution § 8.1(4), p. 383 (2d ed.1993); or to estimate how much it would cost the victim (if he survived) to obtain counseling or therapy to minimize the pain and suffering, Law Commission, Damages for Personal Injury: Non Pecuniary Loss 8 (Consultation Paper No. 140, 1995); Andrews v. Grand & Toy Alberta Ltd., (1978) 83 D.L.R. (3d) 452, 476-77 (Can.S.Ct.), or how much they would demand to assume the risk of the pain and suffering that the victim experienced. Mark Geistfeld, “Placing a Price on Pain and Suffering: A Method for Helping Juries Determine Tort Damages for Nonmonetary Injuries,” 83 Cal. L.Rev. 773, 818-28 (1995). If they said they would demand $1,000 to assume a .01 risk of such a misfortune, this would imply that the victim should receive an award of $100,000, as that is the judgment that, if anticipated, would have induced the defendant to spend up to $1,000 to prevent. Talk is cheap, though; and maybe a *759better approach would be to present the jury with evidence of how potential victims themselves evaluate such risks, an approach that has been used to infer the value of life from people’s behavior in using safety devices such as automobile seat-belts or in demanding risk premiums to work at hazardous jobs. See, e.g., Richard Thaler & Sherwin Rosen, “The Value of Saving a Life: Evidence from the Labor Market,” in Household Production and Consumption 265 (Nestor E. Terleckyj ed.1975); W. Kip Viscusi, “The Value of Risks to Life and Health,” 31 J. Econ. Lit.1912 (1992); Paul Lanoie, Carmen Pedro & Robert Latour, “The Value of a Statistical Life: A Comparison of Two Approaches,” 10 J. Risk & Uncertainty 235 (1995).
Most courts do not follow any of these approaches. Instead they treat the determination of how much damages for pain and suffering to award as a standardless, unguided exercise of discretion by the trier of fact, renewable for abuse of discretion pursuant to no standard to guide the reviewing court either. To minimize the arbitrary variance in awards bound to result from such a throw-up-the-hands approach, the trier of fact should, as is done routinely in England, J. Munkman, Damages for Personal Injury and Death 162-63 (7th ed.1985); Law Commission, supra, at 19-22, be informed of the amounts of pain and suffering damages awarded in similar cases. See, e.g., Johnson v. Rogers & Phillips, Inc., 753 So.2d 286, 295 (La.App. 1999); Oscar G. Chase, “Helping Jurors Determine Pain and Suffering Awards,” in Reforming the Civil Justice System 340 (Larry Kramer ed.1996); James F. Blumstein, Randall R. Bovbjerg & Frank A. Sloan, “Beyond Tort Reform: Developing Better Tools for Assessing Damages for Personal Injury,” 8 Yale J. Reg. 171, 184-85 (1991). And when the trier of fact is a judge, he should be required as part of his Rule 52(a) obligation to set forth in his opinion the damages awards that he considered comparable. We make such comparisons routinely in reviewing pain and suffering awards, see, e.g., Tullis v. Townley Engineering & Mfg. Co., 243 F.3d 1058, 1066 (7th Cir.2001); Riemer v. Illinois Dept. of Transportation, 148 F.3d 800, 808 (7th Cir.1998); DeBiasio v. Illinois Central R.R., 52 F.3d 678, 689 (7th Cir.1995), as do other courts. See, e.g., Smith v. Kmart Corp., 177 F.3d 19, 31-32 (1st Cir.1999); Mathie v. Fries, 121 F.3d 808, 813-14 (2d Cir.1997); Williams v. Chevron U.S.A., Inc., 875 F.2d 501, 506 (5th Cir.1989). It would be a wise practice to follow at the trial level as well.
But can we prescribe it in a case such as this where the rule of decision is state rather than federal? The Supreme Court has held that a rule requiring uniformity in damages awards is, like a cap on damages, see Knowles v. United States, 91 F.3d 1147, 1150 (8th Cir.1996); Aguilar v. United States, 920 F.2d 1475, 1478 (9th Cir.1990); Reilly v. United States, 863 F.2d 149, 161 (1st Cir.1988), substantive within the meaning of the Erie doctrine. Gasperini v. Center for Humanities, Inc., 518 U.S. 415, 428-31, 116 S.Ct. 2211, 135 L.Ed.2d 659 (1996). But the Erie doctrine, a limitation on federal judicial intervention in disputes that arise under state law, is not directly applicable to a suit against the United States. Here state law is being borrowed only because there is no well developed body of federal common law concerning the common torts such as negligence, not because there is a state interest in the outcome of the suit. It is true that Erie was not merely a constitutional decision but also an interpretation of the Rules of Decision Act, 28 U.S.C. § 1652 — but that Act provides only that “the laws of the several states ... shall be regarded as rules of decision in civil actions in the courts of the United States, in cases where they [the laws of the several states] apply” (emphasis added), and they *760do not apply, in the sense of govern, the conduct of the United States.
The distinction between borrowing and applying state law is perhaps most clearly established in cases involving the borrowing of a state statute of limitations for use in litigation based on a federal statute. See, e.g., West v. Conrail, 481 U.S. 35, 39 n. 4, 107 S.Ct. 1538, 95 L.Ed.2d 32 (1987); Doe v. Blue Cross & Blue Shield United of Wisconsin, 112 F.3d 869, 873-74 (7th Cir. 1997); McIntosh v. Antonino, 71 F.3d 29, 36 (1st Cir.1995); Hemmings v. Barian, 822 F.2d 688, 689-90 (7th Cir.1987). As we put it in Hemmings, “When a federal court borrows a state statute of limitations for use in connection with a federal statute that does not have its own statute of limitations, the court is not applying state law; it is applying federal law. It looks to state law for guidance.... The analysis would, however, be different if Count I were a diversity rather than federal question count. For purposes of the Erie doctrine the statute of limitations is substantive rather than procedural, and the federal court therefore applies state law-it doesn’t just borrow it.” Id.
Nor is a practice of consulting damages awards in comparable cases for purposes of facilitating a more thoughtful, disciplined, and informed award in the particular case the same thing as a rule limiting awards within a range set by previous cases, one understanding of the state law rule • involved in Gaspeñni: an award of damages was not to “deviate materially” from awards that had been made in comparable cases. Nor is it easy to characterize a practice of not consulting comparable awards as a remedial rule, hence “substantive” in the Erie sense, rather than as a rule of evidence; .or a. practice of such consultation by a reviewing court endeavoring to carry out its duty to prevent abuses of discretion in the award of damages as anything more than a rule of appellate procedure. So on multiple grounds, what indeed appears to be the rule in Illinois, that comparable awards for pain and suffering are not to be considered at either the appellate or the trial stage, see, e.g., Epping v. Commonwealth Edison Co., 315 Ill.App.3d 1069, 248 Ill.Dec. 625, 734 N.E.2d 916, 918-19 (Ill.App.2000); Barry v. Owens-Corning Fiberglas Corp., 282 Ill.App.3d 199, 217 Ill.Dec. 823, 668 N.E.2d 8, 14 (Ill.App.1996); Simmons v. University of Chicago Hospitals & Clinics, 247 Ill.App.3d 177, 187 Ill.Dec. 70, 617 N.E.2d 278, 288 (Ill.App.1993), does not bind us in this case.
In any event, any argument that we should not look at comparable awards in reviewing the award of pain and suffering damages in this case has been forfeited by the plaintiff, who, while citing Epping for the proposition stated in her brief that “not all courts agree that comparison of damage awards aids in evaluating their propriety,” has not argued that Illinois forbids such comparisons or that, if it does, its rule binds us. Instead, both parties have cited what they deem comparable cases. Only their notions of comparability are stunted. The plaintiff cites three cases in which damages for pain and suffering ranging from $600,000 to $1 million were awarded, but in each one the pain and suffering continued for hours, not minutes. The defendant confined its search for comparable cases to other pñson suicide cases, implying that prisoners experience pain and suffering differently from other persons, so that it makes more sense to compare Johnson’s pain and suffering to that of a prisoner who suffered a toothache than to that of a free person who was strangled, and concluding absurdly that any award for pain and suffering in this case that exceeded $5,000 would be excessive. The parties should have looked at awards in other cases involving asphyxiation, for example cases of drowning, which are numerous. See, e.g., Millman v. *761County of Butler, 244 Neb. 125, 504 N.W.2d 820, 826 (Neb.1993); Snyder v. Whittaker Corp., 839 F.2d 1085, 1092-93 (5th Cir.1988); Dontas v. City of New York, 183 A.D.2d 868, 584 N.Y.S.2d 134 (App.Div.1992) (per curiam); Turner v. Parish of Jefferson, 721 So.2d 64, 76-78 (La.Ct.App.1999); Stissi v. Interstate & Ocean Transport Co., 590 F.Supp. 1043, 1048-49 (E.D.N.Y.1984), aff'd. in relevant part, 765 F.2d 370, 377 (2d Cir.1985). Had they done so, they would have come up with an award in the range of $15,000 to $150,000.
There is more wrong with the award. The $1.6 million awarded by the judge was far higher than the amount requested by the plaintiff (which was $300,000 to $600,000) and came after the judge, in accordance with the principle of comparative negligence, had reduced the award on the ground that Johnson’s own conduct, that is, his hanging himself, required his estate to bear some of the responsibility for the pain and suffering that the hanging inflicted on him. The judge did not indicate from what level (or by what percentage) he had reduced the award to $1.6 million.
But as we said the plaintiff failed to establish liability. The judgment must therefore be reversed with instructions to enter judgment for the government.
REVERSED.