with whom Justice Marshall and Justice Blackmun join, concurring in part and concurring in the judgment.
I agree that the “single incident” instruction, ante, at 813, is properly before us and therefore join Part II of Justice *825Rehnquist’s opinion. Although I concur in the judgment reached by the Court today, I am unable to join the balance of the plurality opinion.
Monell v. New York City Dept. of Social Services, 436 U. S. 658 (1978), held that municipalities, like other state actors, are subject to liability under § 1983 when their policies “subjec[t], or caus[e] to be subjected, any citizen of the United States ... to the deprivation of any rights, privileges, or immunities secured by the Constitution . . . .” 42 U. S. C. § 1983. I agree with the plurality that today we must take a “small but necessary step,” ante, at 810, toward defining the full contours of municipal liability pursuant to § 1983.1 However, because I believe that the plurality opinion needlessly complicates this task and in the process unsettles more than it clarifies, I write separately to suggest a simpler explanation of our result.
h-H
Given the result m this case, in which a jury verdict in favor of the respondent is overturned, it is useful to keep in mind respondent’s theory of the case. Respondent introduced two types of evidence at trial. First, respondent elicited testimony concerning the circumstances surrounding Tuttle’s killing. This included Rotramel’s admission that he never saw a weapon in Tuttle’s possession, App. 150, 158, 225, and evidence that there was no reasonable ground to believe that Tuttle had committed a felony. Id., at 155.2 *826It also included evidence that Rotramel made no effort to employ alternative measures to apprehend Tuttle, id., at 225-226. Second, respondent introduced substantial direct evidence concerning what she alleged to be the city’s grossly inadequate policies of training and supervising police officers. An expert testified that Rotramel’s training included only 24 minutes of instruction in how to answer calls concerning a robbery in progress, although “these are statistically one of the most dangerous calls that an officer has to handle.” Id., at 288. In addition, there was evidence that Rotramel had little or no training in when or how to enter a “blind” building with an armed robbery in progress and whether to wait for a backup unit to arrive. Id., at 146-147. Finally, Rotramel himself seemed to believe that he had been inadequately trained. Id., at 153, 159, 165.
Respondent thus attempted in two ways to show the city’s responsibility for the killing of Tuttle. First, respondent proposed to prove that Rotramel’s killing of Tuttle was so egregiously out of accord with accepted police practice that the jury could infer from the killing alone that the city’s policies and customs concerning the training of police were grossly deficient and were to blame for the incident. Second, respondent hoped to prove the policy or custom of inadequate training by means of direct evidence of the scope and nature of that training.
The trial court permitted respondent to submit both theories to the jury. The jury was instructed that “a single, unusually excessive use of force may be sufficiently out of the ordinary to warrant an inference that it was attributable to inadequate training or supervision amounting to ‘deliberate indifference’ or ‘gross negligence’ on the part of the officials in charge.” Id., at 44. The court had previously instructed that “deliberate indifference” or “gross negligence” on the part of the city was sufficient to prove the existence of a city policy. Id., at 43. Putting these instructions together, the *827jury could infer solely from evidence concerning the conduct of a single policeman on a single night that the city was liable under § 1983. As for the second theory, the jury was instructed that the city could be held liable “only if an official policy which results in constitutional deprivations can be inferred from acts or omissions of supervisory city officials and if that policy was a proximate cause of the denial of the civil rights of the decedent.” Ibid.
Having been thus instructed, the jury returned a verdict against the city. There is no way to determine on which theory the jury relied. The trial court denied petitioner’s motion for judgment notwithstanding the verdict, holding that “the plaintiff brought forward sufficient evidence regarding inadequate training and procedures to warrant submission to the jury of the issue of municipal liability.” Id., at 58. The court believed that “there was considerably more evidence presented here than the fact that [Rotramel], a young man, shot someone in deprivation of their civil rights.” Tr. 704. In discussing petitioner’s judgment n.o.v. motion, the court explicitly noted that it was “impressed with the evidence that was presented in this case” concerning “the curriculum methods and the lack of supervision and training.” Id., at 704-705. The Court of Appeals affirmed. 728 F. 2d 456 (CA10 1984).
The question presented in the petition for certiorari is “[w]hether a single isolated incident of the use of excessive force by a police officer establishes an official policy or practice of a municipality sufficient to render the municipality liable for damages under 42 U. S. C. § 1983.” The thrust of petitioner’s argument is that it was improper to instruct the jury that it could impose liability on petitioner based solely on evidence regarding Rotramel’s actions on the night of Tuttle’s killing.
II
A
Monell v. New York City Dept. of Social Services, 436 U. S. 658 (1978), held that “Congress did intend municipal*828ities and other local government units to be included among those persons to whom § 1983 applies.” Id., at 690 (emphasis in original). Nonetheless, we recognized certain limits on the theories of liability that could be asserted against a municipality. As the plurality correctly notes, ante, at 817-818, our reading in Monell of the legislative history of § 1983, including its rejection of the Sherman amendment, see 436 U. S., at 664-704, led us to conclude that Congress desired not to subject municipalities to liability “without regard to whether a local government was in any way at fault for the breach of the peace for which it was to be held for damages.” Id., at 681, n. 40. We therefore concluded that a city could not be held liable under a vicarious liability or respondeat superior theory in a § 1983 suit, for such liability would violate the evident congressional intent to preclude municipal liability in cases in which the city itself was not at fault.
Because Congress intended that § 1983 be broadly available to compensate individuals for violations of constitutional rights, see Owen v. City of Independence, 445 U. S. 622, 650-653 (1980); Monell, supra, at 683-687, a municipality could be held liable where a plaintiff could show that it was the city itself that was at fault for the damage suffered. To make this showing, a plaintiff must prove, in the broad causal language of the statute, that a policy or custom of the city “subjected” him, or “caused him to be subjected” to the deprivation of constitutional rights. In a case in which the plaintiff carries this burden, the city’s liability would be mandated by the language, the legislative history, and the underlying purposes of § 1983.
B
I agree with the plurality that it is useful to begin with the terms of the statute:
“Every person who, under color of any statute, ordinance, regulation, custom, or usage, of any State . . . , subjects, or causes to be subjected, any citizen of the United States or other person within the jurisdiction *829thereof to the deprivation of any rights, privileges, or immunities secured by the Constitution and laws, shall be liable to the party injured in an action at law . . . .”
In the language of the statute, the elements of a § 1983 cause of action might be summarized as follows: The plaintiff must prove that (1) a person (2) acting under color of state law (3) subjected the plaintiff or caused the plaintiff to be subjected (4) to the deprivation of a right secured by the Constitution or laws of the United States. Element (4) involves the question of whether there has been a violation of the Constitution or laws of the United States; that issue is not raised by the parties in this case and thus may be ignored here.
Of the three remaining elements of a § 1983 cause of action of relevance here, respondent clearly established two. After Monell, a municipality like Oklahoma City undoubtedly is a “person” to whom §1983 applies. And there can be little doubt that the city’s actions establishing particular police training procedures were actions taken “under color of state law,” as that term is commonly understood.
The remaining question is causation. In a § 1983 case involving a municipal defendant, the causation element to be proved by the plaintiff may be seen as divided into two parts. First, the plaintiff must predicate his recovery on some particular action taken by the city, as opposed to an action taken unilaterally by a nonpolicymaking municipal employee. This is the inquiry required by Monell, and the plaintiff would carry his burden by proving the existence of a particular official municipal policy or established custom.3 In this case, the municipal policies involved were the set of procedures for training and supervising police officers.4 Second, the plain*830tiff must prove that this policy or custom “subjected” or “caused him to be subjected” to a deprivation of a constitutional right.
The instruction in question in this case permitted the plaintiff to carry his burden of proving “policy or custom” by merely introducing evidence concerning the particular actions taken by Rotramel on the night of October 4, 1980.5 To isolate the defect in this instruction, it is useful to assume that the jury disbelieved Rotramel’s testimony concerning the inadequacy of his training, rejected the evidence presented by respondent’s expert concerning the content of the city’s police training and supervision practices, and found unconvincing all of respondent’s independent and documentary evidence concerning those practices. While perhaps unlikely, such disbelief must be assumed to test an instruction that might have permitted liability without any such evidence. Under the instruction in question, the jury could have found the city liable solely because Rotramel’s actions on the night in question were so excessive and out of the ordinary.
A jury finding of liability based on this theory would unduly threaten petitioner’s immunity from respondeat superior liability. A single police officer may grossly, outrageously, and recklessly misbehave in the course of a single incident. Such misbehavior may in a given case be fairly *831attributable to various municipal policies or customs, either those that authorized the police officer so to act or those that did not authorize but nonetheless were the “moving force,” Polk County v. Dodson, 454 U. S. 312, 326 (1981), or cause of the violation. In such a case, the city would be at fault for the constitutional violation. Yet it is equally likely that the misbehavior was attributable to numerous other factors for which the city may not be responsible; the police officer’s own unbalanced mental state is the most obvious example. Cf. Brandon v. Holt, 469 U. S. 464, 466 (1985). In such a case, the city itself may well not bear any part of the fault for the incident; there may have been nothing that the city could have done to avoid it. Thus, without some evidence of municipal policy or custom independent of the police officer’s misconduct, there is no way of knowing whether the city is at fault. To infer the existence of a city policy from the isolated misconduct of a single, low-level officer, and then to hold the city liable on the basis of that policy, would amount to permitting precisely the theory of strict respondeat superior liability rejected in Monell.6
Respondent objects that in Monell and Owen v. City of Independence, 445 U. S. 622 (1980), we found a municipality liable despite evidence that showed only a single instance of misconduct. If the city’s argument here depended on the premise that municipal conduct that resulted in only a single *832incident was immune from liability, I would have to agree with respondent that Monell and Owen provide authority to the contrary. A rule that the city should be entitled to its first constitutional violation without incurring liability — even where the first incident was the taking of the life of an innocent citizen — would be a legal anomaly, unsupported by the legislative history or policies underlying § 1983. A § 1983 cause of action is as available for the first victim of a policy or custom that would foreseeably and avoidably cause an individual to be subjected to deprivation of a constitutional right as it is for the second and subsequent victims; by exposing a municipal defendant to liability on the occurrence of the first incident, it is hoped that future incidents will not occur.
The city’s argument, however, does not depend on any such unlikely or extravagant premise. It depends instead merely on that fact that a single incident of police misbehavior by a single policeman is insufficient as sole support for an inference that a municipal policy or custom caused the incident. And this was not an inference comparable to any on which the plaintiffs in Monell or Owen relied. In Monell, both parties agreed that the City of New York had a policy of forcing women to take maternity leave before medically necessary. 436 U. S., at 661, n. 2. This policy, of course, violated the interest we recognized in Cleveland Board of Education v. LaFleur, 414 U. S. 632 (1974). In Owen, the municipality’s city council, in the course of dismissing the plaintiff from his post as Chief of Police, passed a resolution releasing to the press material that smeared the reputation of the plaintiff. There was no doubt that the release of the information was an official action — that is, a policy or custom— of the city. Thus, the crucial factor in both cases was that the plaintiff introduced direct evidence that the city itself had acted.7 In both cases, the jury was not required to draw any *833further inference concerning the existence of the city policy, let alone an inference from the isolated conduct of a single nonpolicymaking city employee on a single occasion.8
HH HH
For the reasons given above, I agree with the Court that the judgment in this case should be reversed; there may be many ways of proving the existence of a municipal policy or custom that can cause a deprivation of a constitutional right, but the scope of § 1983 liability does not permit such liability to be imposed merely on evidence of the wrongful actions of a single city employee not authorized to make city policy.9
See Monell v. New York City Dept. of Social Services, 436 U. S. 658, 695 (1978). Since Monell, of course, the contours of municipal liability have become substantially clearer. See, e. g., Newport v. Fact Concerts, Inc., 453 U. S. 247 (1981) (punitive damages not permitted); Owen v. City of Independence, 445 U. S. 622 (1980) (qualified immunity not available to municipalities).
Rotramel himself admitted at the time he entered the bar, Tuttle was standing with a drink in his hand. App. 155. There was also testimony that the bartender told Rotramel that no robbery had occurred, id., at *82682-83, 106, 234, and Rotramel conceded that no one in the bar told him that a robbery had occurred. Id., at 209.
Of course, nothing hinges on whether the “policy or custom” inquiry is seen as a part of the plaintiff’s burden to prove causation, or whether instead it is seen as an independent element of a § 1983 cause of action.
These included official decisions concerning the following matters: whether to permit rookie police officers to patrol alone; what rules should govern whether a police officer should wait for backup units before entering a felony-in-progress situation; how much time and emphasis to be *830placed on training in such matters as how to approach felony-in-progress situations, when to use firearms, and when to shoot to kill. Respondent bore the burden at trial of proving that the alleged deprivation of constitutional rights (the killing of Tuttle) resulted from these “conscious choices,” ante, at 823, made by the city concerning police training and supervision.
Rotramel was a low-level police officer. Some officials, of course, may occupy sufficiently high policymaking roles that any action they take under color of state law will be deemed official policy. See Monell, 436 U. S., at 694 (“[I]t is when execution of a government’s policy or custom, whether made by its lawmakers or by those whose edicts or acts may fairly be said to represent official policy, inflicts the injury that the government as an entity is responsible under § 1983”).
This is in some respects analogous to the doctrine of res ipsa loquitur in ordinary tort cases. Only in certain circumstances in ordinary tort cases may a jury infer defendant’s fault from the fact that an injury of a certain type occurred. See generally W. Keeton, D. Dobbs, R. Keeton, & D. Owen, Prosser and Keeton on Law of Torts § 39, p. 243 (5th ed. 1984). The purpose of the restriction is of course to protect the defendant from liability in a ease in which he is not at fault and has not caused the injury. The jury instruction in question here similarly would have permitted the city to be held liable, absent fault and causation. This suggests that there may be cases, analogous to those in which the res ipsa loquitur doctrine applies, where the evidence surrounding a given incident is sufficient to permit a jury to infer that it was caused by a city policy or custom.
The distinction between Monell and Owen, on the one hand, and the instant ease, on the other, is thus rather simple. In Monell and Owen, *833the plaintiff introduced evidence of official actions taken by the defendant municipality. Respondent here, of course, also introduced evidence concerning official actions taken by the city, mostly centering on the city policies governing training and supervision of police officers. However, as the plurality points out, ante, at 821-822, the judgment must be reversed in this case because the instructions permitted the jury to find the city liable even if the jury did not believe this direct evidence. Cf. Stromberg v. California, 283 U. S. 359, 367-368 (1931).
I do not understand, nor do I see the necessity for, the metaphysical distinction between policies that are themselves unconstitutional and those that cause constitutional violations. See ante, at 823-824, and n. 7. If a municipality takes actions — whether they be of the type alleged in Monell, Owen, or this case — that cause the deprivation of a citizen’s constitutional rights, § 1983 is available as a remedy.
The plurality seems to believe that there is a serious threat that a court might submit to a jury the theory that a municipal policy of having a police department was the “cause” of a deprivation of a constitutional right. Ante, at 823. Of course, I agree that such a theory should never be submitted to a jury, but the reason has little to do with the presence of the municipality as the defendant in the case or the structure of § 1983. Ordinary principles of causation used throughout the law of torts recognize that “but for” causation, while probably a necessary condition for liability, is never a sufficient condition of liability. See generally Prosser & Keeton on Law of Torts § 41, at 265-266. I would think that these principles are sufficient to avoid the unusual theory of liability suggested by the plurality.