Anthony Dye was injured while attempting to flee from the police in Elkhart, Indiana. After his capture, Dye pleaded guilty to three state felonies he committed during these events: attempted battery with a deadly weapon (a charge reduced from attempted murder), possession of a handgun by a convicted felon, and possessing a handgun within 1,000 feet of a school. In this federal litigation under 42 U.S.C. sec.1983 Dye seeks to turn the tables and collect damages on account of the injuries he sustained while being subdued. But the district court granted summary judgment in defendants' favor, ruling that Dye's claims are covered by a release.

Dye was carrying a firearm while driving his Corvette at 2:30 a.m. one day in Elkhart, Indiana. Officer William Wargo pulled behind the Corvette with his flashing lights on. Dye did not stop and made for his mother's house. After pulling into the driveway, Dye leapt from the car and ran toward the door. Wargo told his K-9, Frei, to prevent Dye's escape, which Frei did by biting one of Dye's legs, as Frei had been trained to do. At Wargo's direction, Dye assumed a prone position, and Frei released his leg. Before he could be handcuffed, however, Dye got up, pulled a semi- automatic pistol from his waistband, and opened fire. Wargo returned fire and called on Frei for aid. Dye got the worst of things: against Dye's multiple gunshot wounds (to his chest and both legs) and dog bites, Wargo suffered only a pinched nerve in his neck. Frei later received awards for valor in the line of duty.

Although this much is common ground, vital details are disputed. Wargo says that Dye was speeding, driving erratically, and ran a stop sign; Dye says that he was obeying all traffic laws. Wargo says that he activated his siren as well as his flashing lights; Dye denies hearing a siren. Dye asserts that he fled because the Elkhart police have a reputation for mistreating young black suspects; Elkhart denies that it has such a reputation. (An alternative hypothesis is that Dye hoped that he could hide the gun in his mother's house and avoid the stiff penalty for possession by a felon. But the reason for his flight is legally irrelevant, and Dye's explanation, even if true, is no justification.) Wargo contends that Frei released Dye after halting his flight and did not attack a second time until Dye refused to be handcuffed and sprang to his feet; Dye contends that the sequence was reversed and that he got back up to defend himself against Frei's unprovoked attack. Dye asserts that he shot at Frei only after Wargo refused to call off his dog; Wargo responds that he rather than Frei was Dye's target. According to Dye, his most serious injuries were received after he had given up, thrown the gun away, was again lying down, and had been rendered helpless by a bullet; according to Wargo, Dye had the gun in his hand and was trying to use it when he received his last wounds. If Dye's version is correct, these injuries at least would be actionable under sec.1983, for shooting a disarmed and passive suspect is a clear example of excessive force in violation of the fourth amendment. But if Wargo's version is correct, Dye has no valid com plaint.

Some of the statements that Dye has made under oath in this litigation are inconsistent with statements he made under oath in state court. For example, Dye's current assertion that he never fired at Wargo is inconsistent with the affirmative answer he gave when asked: "And you shot at an Elkhart City Policeman by the name of William Wargo, Jr.?" His current assertion that he fled toward his mother's home only because he feared violence at the hands of the police is inconsistent with this statement made to the state judge: "And by me knowing at the time I had a gun in my possession, you know, I tried to elude him. And being that I was pretty close to my mother's house, you know, I tried to make it there." One or the other of Dye's stories is perjury. His lawyer contends that Dye was entitled to lie in state court to ensure that the judge accepted the favorable plea bargain, and that we should therefore disregard his earlier sworn statements. That is not a position any judicial system can, or does, tolerate. See, e.g., United States v. Stewart, 198 F.3d 984 (7th Cir. 1999); Hugi v. United States, 164 F.3d 378, 381 (7th Cir. 1999). Cf. Cleveland v. Policy Management Systems Corp., 526 U.S. 795, 806 (1999) (collecting cases from every circuit holding that a litigant is bound by answers given during a deposition, despite a later affidavit contradicting those answers, unless there is a legally valid reason why the deposition answers may be superseded). Although Dye observes that his statements when pleading guilty do not contradict anything he has asserted in this federal case about the last few moments of the encounter, why should these statements be believed when the rest of his story is so questionable? How can any court credit statements made by a litigant such as Dye who has proclaimed his willingness (indeed, asserts an entitlement) to lie under oath whenever deceit serves his interests? But we need not pursue this issue, because Dye cannot prevail even if he is entitled to retract his prior testimony.

Two of the three defendants do not belong in this case. In litigation under sec.1983 a municipality is not vicariously liable for the constitutional torts of its employees but is answerable only for the consequences of its policies. See Monell v. Department of Social Services, 436 U.S. 658 (1978). Elkhart does not have a policy of shooting suspects when they are down. Although Dye contends that Elkhart did not properly train either Wargo or Frei, shortcomings of this kind do not establish direct liability, because the Constitution does not require municipalities to conduct training programs. Poor training is instead a means of showing intent for those constitutional torts where intent matters, see Collins v. Harker Heights, 503 U.S. 115, 122-24 (1992); Canton v. Harris, 489 U.S. 378, 388 (1988), and excessive force under the fourth amendment is not one of those constitutional torts. See Graham v. Connor, 490 U.S. 386 (1989); Lester v. Chicago, 830 F.2d 706 (7th Cir. 1987). Proof of failure to train officers could be used to demonstrate that the municipality approves (hence has a policy of) improper conduct that training could extirpate. Such a claim in a case like this would depend on establishing that the City's policymakers knew that the police were using objectively unreasonable force in apprehending suspects, yet did nothing to solve the problem. See Canton, 489 U.S. at 388 n.8; Lanigan v. East Hazel Crest, 110 F.3d 467, 478-79 (7th Cir. 1997); Sledd v. Lindsay, 102 F.3d 282 (7th Cir. 1996). Dye has not offered any evidence that use of excessive force is common in Elkhart, indeed has not produced evidence of even one prior incident. Cf. Oklahoma City v. Tuttle, 471 U.S. 808, 824 (1985) ("considerably more proof than [a] single incident will be necessary . . . to establish both the requisite fault on the part of the municipality, and the causal connection between the 'policy' and the unconstitutional deprivation"). Thus the City cannot be held liable on the theory that lack of more extensive training for Wargo or Frei evinces a policy of using constitutionally improper force.

As for Frei: sec.1983 applies only to a "person" who acts under color of state law. See Arizonans for Official English v. Arizona, 520 U.S. 43, 69 (1997). Under the Dictionary Act, 1 U.S.C. sec.1, "the words 'person' and 'whoever' include corporations, companies, associations, firms, partnerships, societies, and joint stock companies, as well as individuals", but dogs are not on this list, whether or not they act under color of state law. Cf. Miles v. Augusta City Council, 710 F.2d 1542, 1544 n.5 (11th Cir. 1983) (a cat is not a "person" for purposes of the fourteenth amendment). A suit against a dog poses a host of other problems. Was Frei served with process? Did he retain as his lawyer Lynn E. Kalamaros, who purports to represent all three defendants? Was Frei offered the right of self-representation under 28 U.S.C. sec.1654? What relief does Dye seek from a dog--Frei's awards, perhaps? Could Frei claim qualified immunity? If a reasonable person in the defendant's position would not have understood that what he was doing violated the Constitution, damages are unavailable. See Anderson v. Creighton, 483 U.S. 635 (1987). Must we then ask whether a reasonable dog in Frei's position should have understood that he was violating Dye's constitutional rights? One could half understand pursuing Frei because he is not a party to the release. But at oral argument, when asked why he had named a dog as a defendant, Dye's lawyer replied that he deemed Frei an "employee" of the City and was hoping to hold the City vicariously liable for his deeds. That not only ignores Monell but also scotches any effort to skirt the release--for that document covers all of the City's employees. (Anyway, treating a dog as an "employee" would raise thorny issues under the Fair Labor Standards Act. Should Frei get time-and-a-half for overtime? Cf. Brock v. Cincinnati, 236 F.3d 793 (6th Cir. 2001).) All things considered, it is best to follow the Dictionary Act and hold that a dog is not a proper defendant in litigation under sec.1983. (Dye's state-law claims against Frei fare no better; Indiana requires the victim of a dog bite to sue the dog's owner, not the dog. Burgin ex rel. Akers v. Tolle, 500 N.E.2d 763, 766 (Ind. App. 1986).)

This leaves the claim against Wargo, which as the district court held is barred by Dye's release of "any State or Federal claim or cause of action of any kind whatsoever . . . arising out of the arrest and shooting of Anthony H. Dye, on or about March 22, 1997." The release is unconditional, and if this language were not comprehensive enough an additional two pages go on to close every possible loophole. Dye does not deny that the release, taken at face value, bars this suit (and also requires him to reimburse defendants for their legal fees). Nonetheless, Dye insists, the release is subject to an unstated condition: that he obtain a plea bargain superior to the one he entered. The district court rightly held this argument foreclosed by Indiana's parol evidence rule, see Kruse Classic Auction Co. v. Aetna Casualty & Surety Co., 511 N.E.2d 326, 329 (Ind. App. 1987), as well as by the principle that one party's unilateral expectations do not affect a contract's meaning. See Ruff v. Charter Behavioral Health System of Northwest Indiana, Inc., 699 N.E.2d 1171, 1173-74 (Ind. App. 1998). Dye does not contend that he conveyed this expectation to the City or any of its lawyers. Instead of pointing to an ambiguity in the release or to the parties' mutual understanding of its effect, Dye insists that, because he was seriously injured, there must be some escape hatch. This is nothing but wishful thinking; we would have to pitch not only the release but also the body of Indiana's contract law out the window to accept his view. Dye gave up his right to sue Wargo and the City but received in return a promise by Wargo and the City not to sue him. Although he lacks assets (including insurance) that would have made suit attractive, he also knew (or could have learned from his lawyer) that liability on account of efforts to kill a police officer (or even a police dog) could not be discharged in bankruptcy. See 11 U.S.C. sec.523(a)(6). The mutual release enabled Dye to ensure that he would get a fresh start at the end of his imprisonment. This release is not the sort of apparently irrational act that a court should endeavor to overcome.

Dye tries to get mileage from the fact that his is the only signature on the release. Yet, as the district judge pointed out, the statute of frauds requires the signature only of the party sought to be bound. Consolidation Services, Inc. v. KeyBank N.A., 185 F.3d 817, 819-20 (7th Cir. 1999) (Indiana law); Mehling v. Butois County Farm Bureau Co-Op Ass'n, Inc., 601 N.E.2d 5, 7 (Ind. App. 1992). Cf. In re Vic Supply Co., 227 F.3d 928 (7th Cir. 2000).

What Dye needed to show is that the City did not agree to the release, not simply that the City's agents failed to sign the release. This document is a mutual release, not a unilateral waiver, so its effectiveness depends on the City's assent. Dye asserts that the City did not agree, but the only evidence he offers is the missing signature, which just takes us back to the statute of frauds. Dye would have a good point if, for example, his lawyer drafted the release, Dye signed it, and counsel then sent the document to the City, which ignored the proposal. Such a sequence would demonstrate an offer but not an acceptance. What actually happened is significantly different, however. During the plea negotiations Dye's lawyer placed the prospect of a release on the table as a bargaining chip. The prosecutor responded that he would not offer any concession in exchange for Dye's release of civil claims. Still, at the insistence of Dye's attorney, the prosecutor passed the idea of a mutual release on to Elkhart's City Attorney. Elkhart's legal department then prepared a release, which it transmitted to Dye's lawyer through the prosecutor's office. Dye signed the document exactly as tendered and handed it back to the prosecutor, who returned it to the City. Thus we know that, although Dye brought up the idea, the City found it acceptable and approved its every word. Dye did not make a counteroffer; he signed the document the City tendered. Agreement has been established. (Dye does not contend that the City has failed to keep its part of the bargain, or that the City Attorney lacks actual authority to negotiate agreements of this kind on Elkhart's behalf.)

Thus the release is valid under Indiana law and extinguishes Dye's claims. Still, we must consider Dye's contention that the contract is "unenforceable [because] the interest in its enforcement is outweighed in the circumstances by a [federal] public policy harmed by enforcement of the agreement." Newton v. Rumery, 480 U.S. 386, 392 (1987). The premise of Dye's argument is that his version of events is correct. Federal law prevents police from getting off scot free after shooting helpless suspects, Dye contends. Dye's position ignores the point of a release--which is to avoid the need to decide whose story is to be believed. We cannot just assume that Dye is telling the truth now, and that both Wargo (now) and Dye himself (at the time of his guilty plea) have dissembled. It would be necessary to hold a trial to determine whether Dye's current story is correct. Yet to hold such a trial would be to say in effect that no release of liability under sec.1983 can be enforced, for a release would never avert a hearing on the merits of the plaintiff's claim, and all of the associated expense, even if the state actors prevailed in the end. That would make it harder (if not impossible) for parties to settle their differences without litigation.

Dye did not get cash for his settlement, but he did receive value (avoidance of any debt that might hang over him after prison); his legal position here, however, would apply even to persons who executed releases in exchange for monetary settlements. It is difficult to see how making releases unenforceable could help other persons in Dye's position who might very much want to resolve their disputes, only to be rebuffed by municipalities who would be unwilling to enter agreements that their adversaries could choose to discard. See Pierce v. Atchison, Topeka & Santa Fe Ry., 65 F.3d 562 (7th Cir. 1995). It is equally difficult to see why, if a plaintiff in sec.1983 litigation may settle for a pittance once a suit is on file, the same person may not settle for a pittance before initiating litigation.

Newton, the only case on which Dye relies, offers him little aid, for it enforced a release of liability under sec.1983. See also Evans v. Jeff D., 475 U.S. 717 (1986) (plaintiff may agree to forego attorneys' fees under 42 U.S.C. sec.1988). Newton rejects a contention that releases given in exchange for the dismissal of criminal charges always are unenforceable. All of the Justices assumed that an ordinary mutual release of damages liability could be enforced; the question on the table in Newton was whether using criminal charges to obtain a release of civil liability would give the state too much leverage. The Justices who dissented in Newton expressed a concern that prosecutors would use their charging discretion to induce settlement. Even a weak criminal accusation creates a risk of such magnitude that victims of official misconduct may surrender their right to seek civil redress in order to avoid a small chance of lengthy imprisonment. That is a much more substantial concern than any argument Dye advances--for the prosecutor not only did not dismiss the charges against him but also declared that the civil settlement would not be taken into account in the criminal plea bargaining. Because the majority in Newton held that even a release-for-dismissal bargain is enforceable, a simple mutual release of civil liability poses no problems. And this is an ordinary mutual release. Dye contends that it was his "understanding" that he would receive consideration for the release in the form of a better plea agreement, but, as we observed when discussing the parol-evidence problem, that assertion is not backed up by written evidence--and now we add that it is not backed up even by parol evidence about what the City's agents said to Dye or his lawyer. Free-floating "understandings" are irrelevant to the law of contract, state or federal. If Dye contended that the prosecutor said something to give rise to this "understanding" then there might be an issue worth debating; but self-generated beliefs have no legal consequences.

Dye has not cited, and we have not found, any case holding that a mutual release of civil liability is unenforceable under federal law. Federal courts have not embraced the view, see Owen M. Fiss, Against Settlement, 93 Yale L.J. 1073 (1984), that settlement interferes with judges' ability to declare the law, right wrongs, and otherwise act as ombudsmen. Litigation offers a means to vindicate claims, but entitlement is not compulsion. Section 1983 and associated statutes do not employ the approach of the Fair Labor Standards Act and the handful of other federal laws that either foreclose private settlements or require their supervision by a public official. See 29 U.S.C. sec.216(c). Waivers and releases serve the interests of both parties: a waivable right is more valuable to its holder than is a non-waivable right, for the waivable right may be traded to the other side for a benefit that the holder values more highly than the right's exercise. See, e.g., United States v. Krilich, 159 F.3d 1020 (7th Cir. 1998).

Circumstances amounting to duress, the kind of threats that undermine any contract, would preclude enforcing a release as well. But Dye does not contend that his release was extracted by improper threats or was otherwise involuntary. Recall that Dye himself (through his lawyer) first proposed the release, persisting after the prosecutor said that civil liability would not be taken into account in the plea bargaining process. Federal law allows parties to waive not only claims for damages, as in Newton, but also the rights to defend themselves (as Dye did when pleading guilty) and to appeal from adverse decisions. See United States v. Mezzanatto, 513 U.S. 196 (1995); United States v. Wenger, 58 F.3d 280 (7th Cir. 1995). Dye must keep his promise to refrain from civil suit, just as his plea of guilty precludes most avenues of attacking his conviction. See Bousley v. United States, 523 U.S. 614 (1998); United States v. Broce, 488 U.S. 563 (1989); Mabry v. Johnson, 467 U.S. 504, 508-09 (1984).

DIANE P. WOOD, Circuit Judge, dissenting in part and concurring in part in the judgment.

Both the majority and I agree that Officer William Wargo, of the Elkhart, Indiana police force, used constitutionally excessive force against Anthony Dye, when we take the contested facts in the light most favorable to Dye. Our disagreement centers on the legal effect of the release Dye signed, which the majority thinks lets both the City of Elkhart and its employee off the hook. I believe this conclusion fails to give proper effect to the facts surrounding the execution of the release and the Supreme Court's instructions about the way we are to assess such documents. I therefore respectfully dissent with respect to the case against Wargo.

* Before turning to the release itself, I believe it is necessary to give a somewhat more detailed account of the facts of the encounter between Dye and Wargo. I do so because I think Dye's excessive force claim extends to more conduct than the majority would recognize, because parts of the majority's account either omit critical facts favorable to Dye or dwell on facts that are peripheral, and because the analysis of the release must be undertaken with the context of the dispute in mind. Naturally, my account of the facts presents them in the light most favorable to Dye, the non-moving party.

In his sec. 1983 lawsuit, Dye named as defendants Wargo, the City of Elkhart, and (mysteriously) Frei, Wargo's police dog. Only the claims against the first two defendants require our attention. (The majority hardly needs to belabor the point that, no matter how much of an animal lover one may be, a dog at this time is not a "person" amenable to a sec.1983 suit). Dye asserted several excessive force claims against Wargo, all stemming from the confrontation in the early morning hours of March 22, 1997. Dye's claim against the City of Elkhart is that his injuries resulted from its failure adequately to train Wargo (and Frei) and that this failure to train rose to the level of deliberate indifference to Dye's right under the Fourth Amendment to be free from unreasonable seizures.

As the majority has reported, the events that triggered this lawsuit began around 2:30 a.m. on March 22, 1997. Dye was driving his brother's Chevrolet Corvette on Indiana Avenue, within the City of Elkhart, headed toward his mother's house. He was obeying the speed limit and all other traffic laws when he observed that he was being followed by a police car, which turned out to be driven by Officer Wargo. Dye reached the intersection of Indiana Avenue and Sterling Avenue and, still in compliance with all traffic laws, came to a full stop. He then turned right onto Sterling Avenue. Nonetheless, after he made the turn, he saw that the police car had turned on its flashing overhead lights. Dye realized that the officer was signaling to him to pull over, but he continued driving in the short-sighted hope that he might be able to reach his mother's house and get inside before the police officer could stop him. His motivation was simple: Dye was a convicted felon and he was carrying an unlicenced 9mm handgun. With Wargo now in pursuit, Dye turned into the alley behind his mother's house, drove until he came to her yard, pulled in and stopped the car.

The minute his car came to a halt, Dye opened the door and without looking back made a dash for the house. Wargo, who had pulled in behind him, saw Dye start running toward the house and released Frei. Wargo never ordered Dye to stop, nor did he warn Dye that he was about to release the dog. Frei overtook Dye before Dye reached the house. As he was trained to do, Frei bit Dye's leg and held on. Wargo yelled to Dye that Frei would not release until Dye got down on the ground in a cross position. Dye did as he was told, and Frei released his grip. Wargo then told Dye to put his hands behind his back. As Dye attempted to comply, Frei attacked him again. Fearful and in pain, Dye stood up, attempting to get Frei to quit biting him and yelling to Wargo to call off the dog. Wargo did nothing, and Frei continued biting. Wargo told Dye that Frei would not stop biting him until he laid down again on the ground. Afraid of what the dog would do to him if he laid down again, Dye instead continued to fight the dog off. Wargo then sprayed Dye in the face with pepper spray and struck him in the back of the neck. Neither of these interventions brought Dye down. Instead, Dye lifted his shirt and pulled his gun from his waistband. Wargo yelled at Dye not to do it, but Dye fired at least twice. (He claims he was aiming for the dog, but I agree with the majority that this is beside the point for purposes of assessing Wargo's conduct. Wargo obviously had no way of knowing whom or what Dye meant to be shooting.) Dye's actions prompted Wargo to pull his own weapon. Once Dye began shooting, Wargo dropped to the ground and fired at Dye, striking him just under the left shoulder.

Wargo's initial shot at last caused Dye to fall to the ground and drop his gun. He wound up face down on the ground with Frei still biting at him. Despite the fact that Dye was now unarmed and on his stomach, Wargo continued to fire, pausing at one point to put a new clip in his weapon. An officer who arrived at the scene in the midst of the shooting reported that while he watched, Wargo shot at Dye five or six times from a standing position about ten feet from Dye. Dye suffered multiple gunshot wounds, most of them flesh wounds on the back or sides of his limbs. He had a wound on the rear of his right arm, just below the elbow, as well as on the back of his left arm. Two bullets passed through Dye's right leg and he suffered a flesh wound to his right calf. Wargo suffered only a pinched nerve in his neck, and Frei was unscathed.

Dye identifies four seizures during the course of these events that he contends were unreasonable for constitutional purposes: 1) Wargo's dispatching Frei to capture Dye as he ran toward his mother's house, without any warning or verbal command to Dye to surrender; 2) Frei's attack on Dye as Dye tried to comply with Wargo's command to put his hands behind his head; 3) Wargo's use of the dog, pepper spray, and a hand strike in response to Dye's refusal to get down on the ground; and 4) Wargo's decision to continue firing at Dye as he lay on the ground, face down, and without a weapon. Wargo's first defense to these claims is that the events of that evening did not occur as Dye claims, but this factual dispute cannot be resolved at the summary judgment stage. More productively, Wargo asserts that even if events transpired as Dye says they did, his use of force was at all times objectively reasonable and, to the extent it was not, he is entitled to qualified immunity because at the time of the incident there was no case law clearly establishing that his conduct was unconstitutional.

The majority agrees that if Dye's version of the events is correct, then at least the fourth of these allegations would be actionable under sec. 1983. That much seems indisputable to me. In fact, in my opinion Dye's account of Wargo's actions states at least two excessive force claims for which Wargo would not be entitled to qualified immunity. The first is the one the majority has identified: Wargo's decision to continue shooting at Dye after he was face down on the ground without a weapon. Even if Dye initially fired at Wargo and not the dog, as Dye testified during his plea colloquy in state court, once Dye was down and no longer posed a threat to Wargo, no reasonable police officer in 1997 could believe that he was entitled to continue firing at the backside of an unarmed and disabled individual. In addition, I would find that Dye also has a claim related to the second point he has identified, Frei's unprovoked attack on Dye once Dye had surrendered to Wargo and was attempting to place his hands behind his back. At that point, Dye was under control and was trying to do what Wargo had asked. It has long been well- established that a police officer may not continue to use force against a suspect who is subdued and complying with the officer's orders. See Frazell v. Flanigan, 102 F.3d 877, 884 (7th Cir. 1996) (jury could reasonably conclude that officer who struck subdued suspect in back with nightstick used objectively unreasonable force and was not entitled to qualified immunity); Ellis v. Wynalda, 999 F.2d 243, 247 (7th Cir. 1993) (force that is reasonable while suspect poses threat is no longer reasonable once threat is no longer present); Priester v. Riviera Beach, 208 F.3d 919, 927 (11th Cir. 2000) (denying qualified immunity to officer who in 1994 allowed his dog to attack suspect who was lying on the ground and not resisting). While Frei may not be a "person," he certainly was an instrumentality of force that Wargo was using, and Wargo was responsible for the dog's actions.

The majority concludes, and I agree, that the City of Elkhart is entitled to summary judgment on Dye's failure to train claim. My only difficulty here is with one statement that could be misinterpreted if read out of context. The majority states, ante at 299, that "the Constitution does not require municipalities to conduct training programs." In the abstract, such a statement might be true, but we deal in practicalities rather than abstractions. As the majority properly recognizes, in a case like Dye's, proof of a failure to train could be used to demonstrate an unlawful municipal policy that tolerated the use of excessive force by Elkhart police officers. Ante at 299. Dye's problem here, as the majority points out, is that he had nothing to back up his allegation that the City of Elkhart's K-9 unit training was constitutionally inadequate at the time of his confrontation with Officer Wargo. In fact, what is in the record contradicts Dye's speculations. For example, Dye asserts that there were whole categories of activity for which the City of Elkhart failed to train Frei, including how to apprehend suspects while off a leash, but the training reports offer unrefuted evidence that such activities were part of Wargo and Frei's training. Dye offers no evidence of any other incidents of excessive force similar to the one he allegedly experienced. There is thus no evidence that the City of Elkhart was aware that it had a problem or that its training was not adequately protecting the rights of civilians. I therefore agree with the majority that Dye's failure to train claim cannot succeed.

With this background established, I turn to the release. Like the district court, the majority finds that it is enforceable as a matter of Indiana contract law, and that this is enough to doom Dye's claim. At best, though, enforceability under state law is just the first step in the analysis. It is critical to take into account the fact that Dye executed this waiver of his federal statutory right to sue during plea negotiations with the prosecutor. According to his own testimony and the testimony of the attorney that represented him during the plea negotiations, Dye signed the waiver with the understanding that in exchange he would receive a more favorable plea agreement. These circumstances bring into play the Supreme Court's decision in Town of Newton v. Rumery, 480 U.S. 386 (1987). There the Court held that whether or not a waiver is enforceable is a matter of federal common law and that the salient question is whether enforcing the waiver is consistent with public policy. Because there are disputed issues of fact that pertain to these issues, I would remand this case for further proceedings.

In Rumery, a majority of the Court decided that a release signed by a defendant whose felony witness tampering charge had been dropped in exchange for the release should be enforced. In Part II of the opinion (which did command a Court), Justice Powell wrote:

We begin by noting the source of the law that governs this case. The agreement purported to waive a right to sue conferred by a federal statute. The question whether the policies underlying that statute may in some circumstances render that waiver unenforceable is a question of federal law. We resolve this question by reference to traditional common-law principles, as we have resolved other questions about the principles governing sec. 1983 actions. . . . The relevant principle is well established: a promise is unenforceable if the interest in its enforcement is outweighed in the circumstances by a public policy harmed by enforcement of the agreement.

480 U.S. at 392 (citation omitted). In Part III-A of the opinion, which also garnered the votes of a majority of the Justices, the Court rejected the notion that waiver-release agreements were per se void as against public policy. Instead, Rumery adopted a case-by-case approach which requires courts to assess whether the waiver was entered into voluntarily, whether the prosecutor had a legitimate purpose for entering into the agreement, and whether enforcement of the waiver otherwise furthers the public interest. 480 U.S. at 398.

Rumery's principal holding that waiver- dismissal agreements are not per se unenforceable left many questions unanswered. Decisions from a number of our sister circuits have begun to provide some answers. For example, as Justice O'Connor's separate opinion in Rumery suggested, it is the defendants in a federal civil rights suit who have the burden of proving by a preponderance of the evidence that a waiver was entered into voluntarily, that there was no prosecutorial overreaching, and that the enforcement of the waiver furthers the public interest. See id. at 401 (O'Connor, J.) (concurring in part and concurring in the judgment); Livingstone v. North Belle Vernon Borough, 12 F.3d 1205, 1214 (3d Cir. 1993); Woods v. Rhodes, 994 F.2d 494 (8th Cir. 1993); Lynch v. City of Alhambra, 880 F.2d 1122 (9th Cir. 1989). This means that a district court properly applying the Rumery test cannot grant summary judgment in a release-dismissal case like this one unless it is clear as a matter of law that there are no material issues of fact with respect to these prerequisites to enforceability. Livingstone, 12 F.3d at 1215 (remanding for determination whether there were disputed issues of material fact regarding voluntariness); Woods, 994 F.2d at 500 (finding reasonable minds could not differ on whether release was voluntary or secured by prosecutorial overreaching); Lynch, 880 F.2d at 1129 n.10 (recognizing that "the inquiry that the district court must perform undermines, to some extent, the very purpose of the release-dismissal" but finding that "such an inquiry is necessary to conform with the public policy requirement announced by the Supreme Court in Rumery").

The district court in this case made none of the necessary factual findings or legal determinations required by Rumery. The majority, although it cites Rumery, seems to think that because Rumery rejected the proposition that releases are never enforceable, this must mean that they are always enforceable. Furthermore, the language the majority uses compels the conclusion that it has improperly placed the burden on Dye to show the flaws in the release, instead of putting the burden on the state to show that it meets Rumery's standards. Its two paragraphs discussing Rumery are replete with phrases like "any argument Dye advances," or "Dye has not cited," or "Dye does not contend." As I indicate briefly below, my review of the record convinces me that there are material issues of fact regarding Dye's waiver that preclude granting summary judgment to the defendants on the basis of the waiver.

Whether a criminal defendant voluntarily entered into a waiver-dismissal arrangement depends on the "particular facts and circumstances surrounding [the] case." Livingstone, 12 F.3d at 1211. The majority in Rumery stressed that Rumery was a sophisticated businessman, that he was not in jail when he signed the agreement, that he was represented by counsel, that his counsel drafted the agreement, and that Rumery had three days to consider the deal. The Court also emphasized that Rumery's decision to sign the agreement was "highly rational" because the benefits of the agreement were obvious: "he gained immunity from criminal prosecution in consideration for abandoning a civil suit that he may well have lost." 480 U.S. at 394. Finally, in a footnote, the majority indicated that it would have more confidence in the voluntariness of an agreement if it were presented to the court for approval. Id. at 398 n.10.

The facts of this case present a far more mixed picture. On the one hand, Dye was represented by counsel, the language of the agreement was clear, and he had sufficient time to read and understand it. This evidence favors a finding of voluntariness. On the other hand, Dye was in prison, there is no evidence that he was particularly sophisticated, he did not draft the agreement, the agreement was never presented to the court, and unlike Rumery, who faced a charge with a maximum seven-year sentence, Dye was facing an attempted murder charge. The pressures on Dye were thus considerably greater than those facing Rumery. Moreover, despite the majority's attempt to rationalize the agreement as giving Dye the opportunity to "start fresh" after serving his sentence, a trier of fact might conclude that the benefits to Dye of signing this waiver agreement were illusory. Even on its face, the agreement required Dye to give up his right to bring a sec.1983 suit against Officer Wargo and the City of Elkhart in exchange only for their not bringing state law tort actions against him (claims that would have been economically foolish given Dye's likely judgment-proof status); Dye received no written promise that any charges against him would be dismissed. This hardly seems like a highly rational judgment, and it would make a reasonable jury question the voluntariness of Dye's agreement to the deal.

The majority attempts to allay any concerns raised by the one-sidedness of the waiver by claiming that the prosecutor informed Dye during their plea negotiations that his signing the waiver would have no effect on the plea negotiations. This, however, is a contested fact in the record. There is no dispute that it was Dye's counsel who suggested the idea of signing a waiver in exchange for a better plea bargain. Contrary to the majority's assertion, however, both Dye and his attorney declared under oath that when Dye signed the mutual release it was their understanding, as a result of the plea negotiations with the prosecutor, that the waiver would indeed affect the plea negotiations and that Dye would receive a more favorable plea agreement in exchange for it. The majority seizes on the word "understanding" as a way of dismissing this testimony, but I would not reject it so readily. Nothing says that only written evidence is competent for Rumery purposes to illuminate the course of the negotiations that led to the contested release. And a factual exploration of these negotiations would not raise the specter of a mini-trial on the underlying lawsuit between Dye and Wargo that the release was designed to avert. It is reasonable to infer from the testimony Dye and his lawyer offered that the prosecutor led Dye to believe that he would receive a more favorable plea agreement if he signed the release. There is no dispute that Dye received no such benefit.

Under traditional contract law, the fact that one party made promises during negotiations that later were not reflected in the plain language of the contract would not be a basis for voiding the contract on voluntariness grounds. But again, the question here is not whether the waiver is valid as a matter of contract law, but rather whether, as a matter of public policy, it should be enforced. For purposes of this analysis, as Rumery makes clear, the question of voluntariness is akin to the standards of "voluntary and knowing" in plea negotiations. And in the plea context, as Fed. R. Crim. P. 11(d) takes pains to emphasize, a court is not entitled to accept any plea of guilty until it determines that the plea was voluntary. Enforceability of a plea agreement under state contract law is entirely beside the point.

In upholding the agreement before it, Rumery also relied on the fact that "the prosecutor had an independent, legitimate reason to make this agreement directly related to his prosecutorial responsibilities." Id. at 398. This finding was critical because all the Justices that joined the majority (and even more so the four dissenting Justices) recognized the risk that release-dismissal agreements could be abused by prosecutors seeking to protect public officials from civil liability. As Justice O'Connor explained in her concurrence, the availability of release agreements may tempt public officials to trump up charges in order to avoid meritorious civil claims, or tempt them "to ignore their public duty by dropping meritorious criminal prosecutions in order to avoid the risk, expense, and publicity of a sec.1983 suit." Id. at 400.

In this case, Wargo's only evidence of a "legitimate reason to make [the waiver] agreement directly related to prosecutorial responsibilities" is that the idea of a waiver was Dye's and that the prosecutor told Dye that it would not be considered as part of the prosecutor's charging decision. Again, however, the latter claim is disputed by the testimony of both Dye and the attorney that represented him in the plea negotiation with the prosecutor. Reading the record in the light most favorable to Dye, we have before us a case where the prosecutor encouraged Dye to sign a waiver agreement with the understanding that his having done so would improve his plea bargain, but that the prosecutor then went back on that verbal promise. The only reason why a prosecutor would use such a strategy is to induce a vulnerable defendant to sign a waiver that would shield public officials from future liability while at the same time not giving up any discretion to prosecute. This is neither a legitimate purpose nor one directly related to prosecutorial responsibilities. Instead, it smacks of bad faith negotiations at best, fraud in the inducement to contract at worst. A waiver obtained by this route cannot, as a matter of public policy, be enforced.

Rumery recognizes that there is a substantial public interest in using sec.1983 to expose and punish unconstitutional conduct by public officials. There is also a substantial interest in not approving practices that have a tendency to undermine the integrity of the criminal justice system. On the other hand, the public has an interest in avoiding frivolous civil litigation, and there will be situations in which entering into a release- dismissal agreement will make sense both from the defendant's point of view and from the point of view of prosecutors with limited resources and other strategic concerns.

If the undisputed facts showed that Dye freely executed this release; if the release had been presented to the court and everyone's expectations about it had been clear; if there was no dispute about the central fact of the scope of the prosecutor's promise (i.e. whether the release would affect the criminal charges or not), I could agree with the majority's disposition of this case. And these counter-factuals show that there certainly will be cases that satisfy the Rumery standards--possibly many, if both sides take care to make an adequate record when they wish to use release- dismissal agreements. But the case I have described is not Dye's case. I would remand this part of the case to the district court for a full factual exploration of the circumstances surrounding the release. After that, it might be possible to dispose of matters at a second round of summary judgment motions, or a trial might be necessary. I respectfully dissent from this part of the judgment.