511 U.S. 661 (1994)

WATERS et al.
v.
CHURCHILL et al.

No. 92-1450.

United States Supreme Court.

Argued December 1, 1993.
Decided May 31, 1994.
CERTIORARI TO THE UNITED STATES COURT OF APPEALS FOR THE SEVENTH CIRCUIT

*663 *663 O'Connor, J., announced the judgment of the Court and delivered an opinion, in which Rehnquist, C. J., and Souter and Ginsburg, JJ., joined. Souter, J., filed a concurring opinion, post, p. 682. Scalia, J., filed an opinion concurring in the judgment, in which Kennedy and Thomas, JJ., joined, post, p. 686. Stevens, J., filed a dissenting opinion, in which Blackmun, J., joined, post, p. 694.

Lawrence A. Manson argued the cause for petitioners. With him on the briefs was Donald J. McNeil.

Richard H. Seamon argued the cause for the United States as amicus curiae urging reversal. With him on the brief were Solicitor General Days, Assistant Attorney General Hunger, Acting Deputy Solicitor General Kneedler, Barbara L. Herwig, and Robert D. Kamenshine.

*664 John H. Bisbee argued the cause for respondents. With him on the brief was Barry Nakell.^[*]

Justice O'Connor announced the judgment of the Court and delivered an opinion, in which The Chief Justice, Justice Souter, and Justice Ginsburg join.

In Connick v. Myers, 461 U. S. 138 (1983), we set forth a test for determining whether speech by a government employee may, consistently with the First Amendment, serve as a basis for disciplining or discharging that employee. In this case, we decide whether the Connick test should be applied to what the government employer thought was said, or to what the trier of fact ultimately determines to have been said.

I

This case arises out of a conversation that respondent Cheryl Churchill had on January 16, 1987, with Melanie Perkins-Graham. Both Churchill and Perkins-Graham were nurses working at McDonough District Hospital; Churchill was in the obstetrics department, and Perkins-Graham was considering transferring to that department. The conversation took place at work during a dinner break. Petitioners heard about it and fired Churchill, allegedly because of it. There is, however, a dispute about what Churchill actually said, and therefore about whether petitioners were constitutionally permitted to fire Churchill for her statements.

*665 The conversation was overheard in part by two other nurses, Mary Lou Ballew and Jean Welty, and by Dr. Thomas Koch, the clinical head of obstetrics. A few days later, Ballew told Cynthia Waters, Churchill's supervisor, about the incident. According to Ballew, Churchill took "`the cross trainee into the kitchen for . . . at least 20 minutes to talk about [Waters] and how bad things are in [obstetrics] in general.' " 977 F. 2d 1114, 1118 (CA7 1992). Ballew said that Churchill's statements led Perkins-Graham to no longer be interested in switching to the department. Supplemental App. of Defendants-Appellees in No. 91-2288 (CA7), p. 60.

Shortly after this, Waters met with Ballew a second time for confirmation of Ballew's initial report. Ballew said that Churchill "was knocking the department" and that "in general [Churchill] was saying what a bad place [obstetrics] is to work." Ballew said she heard Churchill say Waters "was trying to find reasons to fire her." Ballew also said Churchill described a patient complaint for which Waters had supposedly wrongly blamed Churchill. Id., at 67-68.

Waters, together with petitioner Kathleen Davis, the hospital's vice president of nursing, also met with PerkinsGraham, who told them that Churchill "had indeed said unkind and inappropriate negative things about [Waters]." Id., at 228. Also, according to Perkins-Graham, Churchill mentioned a negative evaluation that Waters had given Churchill, which arose out of an incident in which Waters had cited Churchill for an insubordinate remark. Ibid. The evaluation stated that Churchill "`promotes an unpleasant atmosphere and hinders constructive communication and cooperation,' " 977 F. 2d, at 1118, and "`exhibits negative behavior towards [Waters] and [Waters'] leadership through her actions and body language' "; the evaluation said Churchill's work was otherwise satisfactory, id., at 1116. Churchill allegedly told Perkins-Graham that she and Waters had discussed the evaluation, and that Waters "wanted to wipe the slate clean . . . but [Churchill thought] this wasn't possible." *666 Supplemental App. of Defendants-Appellees in No. 91-2288, at 228. Churchill also allegedly told Perkins-Graham "that just in general things were not good in OB and hospital administration was responsible." Id., at 229. Churchill specifically mentioned Davis, saying Davis "was ruining MDH." Ibid. Perkins-Graham told Waters that she knew Davis and Waters "could not tolerate that kind of negativism." Ibid.

Churchill's version of the conversation is different. For several months, Churchill had been concerned about the hospital's "cross-training" policy, under which nurses from one department could work in another when their usual location was overstaffed. Churchill believed this policy threatened patient care because it was designed not to train nurses but to cover staff shortages, and she had complained about this to Davis and Waters. According to Churchill, the conversation with Perkins-Graham primarily concerned the crosstraining policy. 977 F. 2d, at 1118. Churchill denies that she said some of what Ballew and Perkins-Graham allege she said. She does admit she criticized Davis, saying her staffing policies threatened to "ruin" the hospital because they "`seemed to be impeding nursing care.' " Ibid. She claims she actually defended Waters and encouraged Perkins-Graham to transfer to obstetrics. Ibid.

Koch's and Welty's recollections of the conversation match Churchill's. Id., at 1122. Davis and Waters, however, never talked to Koch or Welty about this, and they did not talk to Churchill until the time they told her she was fired. Moreover, Churchill claims, Ballew was biased against Churchill because of an incident in which Ballew apparently made an error and Churchill had to cover for her. Brief for Respondents 9, n. 12.

After she was discharged, Churchill filed an internal grievance. The president of the hospital, petitioner Stephen Hopper, met with Churchill in regard to this and heard her side of the story. App. to Pet. for Cert. 75-77. He then reviewed *667 Waters' and Davis' written reports of their conversations with Ballew and Perkins-Graham, and had Bernice Magin, the hospital's vice president of human resources, interview Ballew one more time. Supplemental App. of Defendants-Appellees in No. 91-2288, at 108, 139-142. After considering all this, Hopper rejected Churchill's grievance.

Churchill then sued under Rev. Stat. § 1979, 42 U. S. C. § 1983, claiming that the firing violated her First Amendment rights because her speech was protected under Connick v. Myers, 461 U. S. 138 (1983). In May 1991, the United States District Court for the Central District of Illinois granted summary judgment to petitioners. The court held that neither version of the conversation was protected under Connick: Regardless of whose story was accepted, the speech was not on a matter of public concern, and even if it was on a matter of public concern, its potential for disruption nonetheless stripped it of First Amendment protection. Therefore, the court held, management could fire Churchill for the conversation with impunity. App. to Pet. for Cert. 45-49.

The United States Court of Appeals for the Seventh Circuit reversed. 977 F. 2d 1114 (1992). The court held that Churchill's speech, viewed in the light most favorable to her, was protected speech under the Connick test: It was on a matter of public concern—"the hospital's [alleged] violation of state nursing regulations as well as the quality and level of nursing care it provides its patients," id., at 1122—and it was not disruptive, id., at 1124.

The court also concluded that the inquiry must turn on what the speech actually was, not on what the employer thought it was. "If the employer chooses to discharge the employee without sufficient knowledge of her protected speech as a result of an inadequate investigation into the employee's conduct," the court held, "the employer runs the risk of eventually being required to remedy any wrongdoing *668 whether it was deliberate or accidental." Id., at 1127 (footnote omitted).

We granted certiorari, 509 U. S. 903 (1993), to resolve a conflict among the Circuits on this issue. Compare the decision below with Atcherson v. Siebenmann, 605 F. 2d 1058 (CA8 1979); Wulf v. Wichita, 883 F. 2d 842 (CA10 1989); Sims v. Metropolitan Dade County, 972 F. 2d 1230 (CA11 1992).

II

A

There is no dispute in this case about when speech by a government employee is protected by the First Amendment: To be protected, the speech must be on a matter of public concern, and the employee's interest in expressing herself on this matter must not be outweighed by any injury the speech could cause to "`the interest of the State, as an employer, in promoting the efficiency of the public services it performs through its employees.' " Connick, supra, at 142 (quoting Pickering v. Board of Ed. of Township High School Dist. 205, Will Cty., 391 U. S. 563, 568 (1968)). It is also agreed that it is the court's task to apply the Connick test to the facts. 461 U. S., at 148, n. 7, and 150, n. 10.

The dispute is over how the factual basis for applying the test—what the speech was, in what tone it was delivered, what the listener's reactions were, see id., at 151-153—is to be determined. Should the court apply the Connick test to the speech as the government employer found it to be, or should it ask the jury to determine the facts for itself? The Court of Appeals held that the employer's factual conclusions were irrelevant, and that the jury should engage in its own factfinding. Petitioners argue that the employer's factual conclusions should be dispositive. Respondents take a middle course: They suggest that the court should accept the employer's factual conclusions, but only if those conclusions were arrived at reasonably, see Brief for Respondents 39, something they say did not happen here.

*669 We agree that it is important to ensure not only that the substantive First Amendment standards are sound, but also that they are applied through reliable procedures. This is why we have often held some procedures—a particular allocation of the burden of proof, a particular quantum of proof, a particular type of appellate review, and so on—to be constitutionally required in proceedings that may penalize protected speech. See Freedman v. Maryland, 380 U. S. 51, 58-60 (1965) (government must bear burden of proving that speech is unprotected); Speiser v. Randall, 357 U. S. 513, 526 (1958) (same); Philadelphia Newspapers, Inc. v. Hepps, 475 U. S. 767, 775-778 (1986) (libel plaintiff must bear burden of proving that speech is false); Masson v. New Yorker Magazine, Inc., 501 U. S. 496, 510 (1991) (actual malice must be proved by clear and convincing evidence); Bose Corp. v. Consumers Union of United States, Inc., 466 U. S. 485, 503-511 (1984) (appellate court must make independent judgment about presence of actual malice).

These cases establish a basic First Amendment principle: Government action based on protected speech may under some circumstances violate the First Amendment even if the government actor honestly believes the speech is unprotected. And though Justice Scalia suggests that this principle be limited to licensing schemes and to "deprivation[s] of the freedom of speech specifically through the judicial process, " post, at 687 (emphasis in original), we do not think the logic of the cases supports such a limitation. Speech can be chilled and punished by administrative action as much as by judicial processes; in no case have we asserted or even implied the contrary. In fact, in Speiser v. Randall, we struck down procedures, on the grounds that they were insufficiently protective of free speech, which involved both administrative and judicial components. Speiser, like this case, dealt with a government decision to deny a speaker certain benefits—in Speiser a tax exemption, in this case a government job—based on what the speaker said. Our *670 holding there did not depend on the deprivation taking place "specifically through the judicial process," and we cannot see how the result could have been any different had the process been entirely administrative, with no judicial review. We cannot sweep aside Speiser and the other cases cited above as easily as Justice Scalia proposes.

Nonetheless, not every procedure that may safeguard protected speech is constitutionally mandated. True, the procedure adopted by the Court of Appeals may lower the chance of protected speech being erroneously punished. A speaker is more protected if she has two opportunities to be vindicated—first by the employer's investigation and then by the jury—than just one. But each procedure involves a different mix of administrative burden, risk of erroneous punishment of protected speech, and risk of erroneous exculpation of unprotected speech. Though the First Amendment creates a strong presumption against punishing protected speech even inadvertently, the balance need not always be struck in that direction. We have never, for instance, required proof beyond a reasonable doubt in civil cases where First Amendment interests are at stake, though such a requirement would protect speech more than the alternative standards would. Compare, e. g., California ex rel. Cooper v. Mitchell Brothers' Santa Ana Theater, 454 U. S. 90, 93 (1981) (per curiam), with McKinney v. Alabama, 424 U. S. 669, 686 (1976) (Brennan, J., concurring in judgment in part). Likewise, the possibility that defamation liability would chill even true speech has not led us to require an actual malice standard in all libel cases. Dun & Bradstreet, Inc. v. Greenmoss Builders, Inc., 472 U. S. 749, 761 (1985) (plurality opinion); Gertz v. Robert Welch, Inc., 418 U. S. 323 (1974). Nor has the possibility that overbroad regulations may chill commercial speech convinced us to extend the overbreadth doctrine into the commercial speech area. Bates v. State Bar of Ariz., 433 U. S. 350, 380-381 (1977).

*671 We have never set forth a general test to determine when a procedural safeguard is required by the First Amendment—just as we have never set forth a general test to determine what constitutes a compelling state interest, see Boos v. Barry, 485 U. S. 312, 324 (1988), or what categories of speech are so lacking in value that they fall outside the protection of the First Amendment, New York v. Ferber, 458 U. S. 747, 763-764 (1982), or many other matters—and we do not purport to do so now. But though we agree with Justice Scalia that the lack of such a test is inconvenient, see post, at 687-688, this does not relieve us of our responsibility to decide the case that is before us today. Both Justice Scalia and we agree that some procedural requirements are mandated by the First Amendment and some are not. See post, at 686. None of us have discovered a general principle to determine where the line is to be drawn. See post, at 686-688. We must therefore reconcile ourselves to answering the question on a case-by-case basis, at least until some workable general rule emerges.

Accordingly, all we say today is that the propriety of a proposed procedure must turn on the particular context in which the question arises—on the cost of the procedure and the relative magnitude and constitutional significance of the risks it would decrease and increase. And to evaluate these factors here we have to return to the issue we dealt with in Connick and in the cases that came before it: What is it about the government's role as employer that gives it a freer hand in regulating the speech of its employees than it has in regulating the speech of the public at large?

B

We have never explicitly answered this question, though we have always assumed that its premise is correct—that the government as employer indeed has far broader powers than does the government as sovereign. See, e. g., Pickering, 391 U. S., at 568; Civil Service Comm'n v. Letter Carri- *672 ers, 413 U. S. 548, 564 (1973); Connick, 461 U. S., at 147. This assumption is amply borne out by considering the practical realities of government employment, and the many situations in which, we believe, most observers would agree that the government must be able to restrict its employees' speech.

To begin with, even many of the most fundamental maxims of our First Amendment jurisprudence cannot reasonably be applied to speech by government employees. The First Amendment demands a tolerance of "verbal tumult, discord, and even offensive utterance," as "necessary side effects of. . . the process of open debate," Cohen v. California, 403 U. S. 15, 24-25 (1971). But we have never expressed doubt that a government employer may bar its employees from using Mr. Cohen's offensive utterance to members of the public or to the people with whom they work. "Under the First Amendment there is no such thing as a false idea," Gertz, supra, at 339; the "fitting remedy for evil counsels is good ones," Whitney v. California, 274 U. S. 357, 375 (1927) (Brandeis, J., concurring). But when an employee counsels her co-workers to do their job in a way with which the public employer disagrees, her managers may tell her to stop, rather than relying on counterspeech. The First Amendment reflects the "profound national commitment to the principle that debate on public issues should be uninhibited, robust, and wide-open." New York Times Co. v. Sullivan, 376 U. S. 254, 270 (1964). But though a private person is perfectly free to uninhibitedly and robustly criticize a state governor's legislative program, we have never suggested that the Constitution bars the governor from firing a high-ranking deputy for doing the same thing. Cf. Branti v. Finkel, 445 U. S. 507, 518 (1980). Even something as close to the core of the First Amendment as participation in political campaigns may be prohibited to government employees. Broadrick v. Oklahoma, 413 U. S. 601 (1973); Letter Carriers, supra; Public Workers v. Mitchell, 330 U. S. 75 (1947).

*673 Government employee speech must be treated differently with regard to procedural requirements as well. For example, speech restrictions must generally precisely define the speech they target. Baggett v. Bullitt, 377 U. S. 360, 367— 368 (1964); Hustler Magazine, Inc. v. Falwell, 485 U. S. 46, 55 (1988). Yet surely a public employer may, consistently with the First Amendment, prohibit its employees from being "rude to customers," a standard almost certainly too vague when applied to the public at large. Cf. Arnett v. Kennedy, 416 U. S. 134, 158-162 (1974) (plurality opinion) (upholding a regulation that allowed discharges for speech that hindered the "efficiency of the service"); id., at 164 (Powell, J., concurring in part and concurring in result in part) (agreeing on this point).

Likewise, we have consistently given greater deference to government predictions of harm used to justify restriction of employee speech than to predictions of harm used to justify restrictions on the speech of the public at large. Few of the examples we have discussed involve tangible, present interference with the agency's operation. The danger in them is mostly speculative. One could make a respectable argument that political activity by government employees is generally not harmful, see Public Workers v. Mitchell, supra, at 99; or that high officials should allow more public dissent by their subordinates, see Connick, supra, at 168-169 (Brennan, J., dissenting); Whistleblower Protection Act of 1989, 103 Stat. 16, or that even in a government workplace the free market of ideas is superior to a command economy. But we have given substantial weight to government employers' reasonable predictions of disruption, even when the speech involved is on a matter of public concern, and even though when the government is acting as sovereign our review of legislative predictions of harm is considerably less deferential. Compare, e. g., Connick, supra, at 151-152; Letter Carriers, supra, at 566-567, with Sable Communications of Cal., Inc. v. FCC, 492 U. S. 115, 129 (1989); Texas v. Johnson, *674 491 U. S. 397, 409 (1989). Similarly, we have refrained from intervening in government employer decisions that are based on speech that is of entirely private concern. Doubtless some such speech is sometimes nondisruptive; doubtless it is sometimes of value to the speakers and the listeners. But we have declined to question government employers' decisions on such matters. Connick, supra, at 146-149.

This does not, of course, show that the First Amendment should play no role in government employment decisions. Government employees are often in the best position to know what ails the agencies for which they work; public debate may gain much from their informed opinions. Pickering, supra, at 572. And a government employee, like any citizen, may have a strong, legitimate interest in speaking out on public matters. In many such situations the government may have to make a substantial showing that the speech is, in fact, likely to be disruptive before it may be punished. See, e. g., Rankin v. McPherson, 483 U. S. 378, 388 (1987); Connick, supra, at 152; Pickering, supra, at 569— 571. Moreover, the government may certainly choose to give additional protections to its employees beyond what is mandated by the First Amendment, out of respect for the values underlying the First Amendment, values central to our social order as well as our legal system. See, e. g., Whistleblower Protection Act of 1989, supra.

But the above examples do show that constitutional review of government employment decisions must rest on different principles than review of speech restraints imposed by the government as sovereign. The restrictions discussed above are allowed not just because the speech interferes with the government's operation. Speech by private people can do the same, but this does not allow the government to suppress it.

Rather, the extra power the government has in this area comes from the nature of the government's mission as employer. Government agencies are charged by law with *675 doing particular tasks. Agencies hire employees to help do those tasks as effectively and efficiently as possible. When someone who is paid a salary so that she will contribute to an agency's effective operation begins to do or say things that detract from the agency's effective operation, the government employer must have some power to restrain her. The reason the governor may, in the example given above, fire the deputy is not that this dismissal would somehow be narrowly tailored to a compelling government interest. It is that the governor and the governor's staff have a job to do, and the governor justifiably feels that a quieter subordinate would allow them to do this job more effectively.

The key to First Amendment analysis of government employment decisions, then, is this: The government's interest in achieving its goals as effectively and efficiently as possible is elevated from a relatively subordinate interest when it acts as sovereign to a significant one when it acts as employer. The government cannot restrict the speech of the public at large just in the name of efficiency. But where the government is employing someone for the very purpose of effectively achieving its goals, such restrictions may well be appropriate.

C

1

The Court of Appeals' decision, we believe, gives insufficient weight to the government's interest in efficient employment decisionmaking. In other First Amendment contexts the need to safeguard possibly protected speech may indeed outweigh the government's efficiency interests. See, e. g., Freedman v. Maryland, 380 U. S. 51 (1965); Speiser v. Randall, 357 U. S., at 526. But where the government is acting as employer, its efficiency concerns should, as we discussed above, be assigned a greater value.

The problem with the Court of Appeals' approach—under which the facts to which the Connick test is applied are determined *676 by the judicial factfinder—is that it would force the government employer to come to its factual conclusions through procedures that substantially mirror the evidentiary rules used in court. The government manager would have to ask not what conclusions she, as an experienced professional, can draw from the circumstances, but rather what conclusions a jury would later draw. If she relies on hearsay, or on what she knows about the accused employee's character, she must be aware that this evidence might not be usable in court. If she knows one party is, in her personal experience, more credible than another, she must realize that the jury will not share that personal experience. If she thinks the alleged offense is so egregious that it is proper to discipline the accused employee even though the evidence is ambiguous, she must consider that a jury might decide the other way.

But employers, public and private, often do rely on hearsay, on past similar conduct, on their personal knowledge of people's credibility, and on other factors that the judicial process ignores. Such reliance may sometimes be the most effective way for the employer to avoid future recurrences of improper and disruptive conduct. What works best in a judicial proceeding may not be appropriate in the employment context. If one employee accuses another of misconduct, it is reasonable for a government manager to credit the allegation more if it is consistent with what the manager knows of the character of the accused. Likewise, a manager may legitimately want to discipline an employee based on complaints by patrons that the employee has been rude, even though these complaints are hearsay.

It is true that these practices involve some risk of erroneously punishing protected speech. The government may certainly choose to adopt other practices, by law or by contract. But we do not believe that the First Amendment requires it to do so. Government employers should be allowed to use personnel procedures that differ from the evidentiary *677 rules used by courts, without fear that these differences will lead to liability.

2

On the other hand, we do not believe that the court must apply the Connick test only to the facts as the employer thought them to be, without considering the reasonableness of the employer's conclusions. Even in situations where courts have recognized the special expertise and special needs of certain decisionmakers, the deference to their conclusions has never been complete. Cf. New Jersey v. T. L. O., 469 U. S. 325, 342-343 (1985); United States v. Leon, 468 U. S. 897, 914 (1984); Universal Camera Corp. v. NLRB, 340 U. S. 474, 490-491 (1951). It is necessary that the decisionmaker reach its conclusion about what was said in good faith, rather than as a pretext; but it does not follow that good faith is sufficient. Justice Scalia is right in saying that we have often held various laws to require only an inquiry into the decisionmaker's intent, see post, at 690-691, but, as discussed supra in Part II—A, this has not been our view of the First Amendment.

We think employer decisionmaking will not be unduly burdened by having courts look to the facts as the employer reasonably found them to be. It may be unreasonable, for example, for the employer to come to a conclusion based on no evidence at all. Likewise, it may be unreasonable for an employer to act based on extremely weak evidence when strong evidence is clearly available—if, for instance, an employee is accused of writing an improper letter to the editor, and instead of just reading the letter, the employer decides what it said based on unreliable hearsay.

If an employment action is based on what an employee supposedly said, and a reasonable supervisor would recognize that there is a substantial likelihood that what was actually said was protected, the manager must tread with a certain amount of care. This need not be the care with which trials, with their rules of evidence and procedure, are *678 conducted. It should, however, be the care that a reasonable manager would use before making an employment decision—discharge, suspension, reprimand, or whatever else— of the sort involved in the particular case. Justice Scalia correctly points out that such care is normally not constitutionally required unless the employee has a protected property interest in her job, post, at 688; see also Board of Regents of State Colleges v. Roth, 408 U. S. 564, 576-578 (1972); but we believe that the possibility of inadvertently punishing someone for exercising her First Amendment rights makes such care necessary.

Of course, there will often be situations in which reasonable employers would disagree about who is to be believed, or how much investigation needs to be done, or how much evidence is needed to come to a particular conclusion. In those situations, many different courses of action will necessarily be reasonable. Only procedures outside the range of what a reasonable manager would use may be condemned as unreasonable.

Petitioners argue that Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274 (1977), forecloses a reasonableness test, and holds instead that the First Amendment was not violated unless "`the defendant[s'] intent [was] to violate the plaintiff[`s] constitutional rights.' " Brief for Petitioners 25; see also post, at 690 (Scalia, J., dissenting). Justice Scalia makes a similar argument based on Pickering, Connick, and Perry, which alluded to the impropriety of management "retaliation" for protected speech. Post, at 689. But in all those cases the employer assertedly knew the true content of the employee's protected speech, and fired the employee in part because of it. In none of them did we have occasion to decide what should happen if the defendants hold an erroneous and unreasonable belief about what plaintiff said. These cases cannot be read as foreclosing an argument that they never dealt with. United States v. L. A. Tucker Truck Lines, Inc., 344 U. S. 33, 38 (1952).

*679 3

We disagree with Justice Stevens' contention that the test we adopt "provides less protection for a fundamental constitutional right than the law ordinarily provides for less exalted rights." Post, at 695. We have never held that it is a violation of the Constitution for a government employer to discharge an employee based on substantively incorrect information. Where an employee has a property interest in her job, the only protection we have found the Constitution gives her is a right to adequate procedure. And an at-will government employee—such as Churchill apparently was, App. to Pet. for Cert. 70—generally has no claim based on the Constitution at all.

Of course, an employee may be able to challenge the substantive accuracy of the employer's factual conclusions under state contract law, or under some state statute or commonlaw cause of action. In some situations, the employee may even have a federal statutory claim. See NLRB v. Burnup & Sims, Inc., 379 U. S. 21 (1964). Likewise, the State or Federal Governments may, if they choose, provide similar protection to people fired because of their speech. But this protection is not mandated by the Constitution.

The one pattern from which our approach does diverge is the broader protection normally given to people in their relationship with the government as sovereign. See, e. g., New York Times Co. v. Sullivan, 376 U. S., at 279-280, cited post, at 696, 699 (Stevens, J., dissenting). But the reasons for this are those discussed supra in Part II—B: "[O]ur `profound national commitment' to the freedom of speech," post, at 699 (Stevens, J., dissenting), must of necessity operate differently when the government acts as employer rather than sovereign.

III

Applying the foregoing to this case, it is clear that if petitioners really did believe Perkins-Graham's and Ballew's *680 story, and fired Churchill because of it, they must win. Their belief, based on the investigation they conducted, would have been entirely reasonable. After getting the initial report from Ballew, who overheard the conversation, Waters and Davis approached and interviewed Perkins-Graham, and then interviewed Ballew again for confirmation. In response to Churchill's grievance, Hopper met directly with Churchill to hear her side of the story, and instructed Magin to interview Ballew one more time. Management can spend only so much of their time on any one employment decision. By the end of the termination process, Hopper, who made the final decision, had the word of two trusted employees, the endorsement of those employees' reliability by three hospital managers, and the benefit of a face-to-face meeting with the employee he fired. With that in hand, a reasonable manager could have concluded that no further time needed to be taken. As respondents themselves point out, "if the belief an employer forms supporting its adverse personnel action is `reasonable,' an employer has no need to investigate further." Brief for Respondents 39.

And under the Connick test, Churchill's speech as reported by Perkins-Graham and Ballew was unprotected. Even if Churchill's criticism of cross-training reported by Perkins-Graham and Ballew was speech on a matter of public concern—something we need not decide—the potential disruptiveness of the speech as reported was enough to outweigh whatever First Amendment value it might have had. According to Ballew, Churchill's speech may have substantially dampened Perkins-Graham's interest in working in obstetrics. Discouraging people from coming to work for a department certainly qualifies as disruption. Moreover, Perkins-Graham perceived Churchill's statements about Waters to be "unkind and inappropriate," and told management that she knew they could not continue to "tolerate that kind of negativism" from Churchill. This is strong evidence that Churchill's complaining, if not dealt with, threatened to undermine *681 management's authority in Perkins-Graham's eyes. And finally, Churchill's statement, as reported by PerkinsGraham, that it "wasn't possible" to "wipe the slate clean" between her and Waters could certainly make management doubt Churchill's future effectiveness. As a matter of law, this potential disruptiveness was enough to outweigh whatever First Amendment value the speech might have had.

This is so even if, as Churchill suggests, Davis and Waters were "[d]eliberately [i]ndifferent," Brief for Respondents 31, to the possibility that much of the rest of the conversation was solely about cross-training. So long as Davis and Waters discharged Churchill only for the part of the speech that was either not on a matter of public concern, or on a matter of public concern but disruptive, it is irrelevant whether the rest of the speech was, unbeknownst to them, both on a matter of public concern and nondisruptive. The Connick test is to be applied to the speech for which Churchill was fired. Cf. Connick, 461 U. S., at 149 (evaluating the disruptiveness of part of plaintiff's speech because that part was "upon a matter of public concern and contributed to [plaintiff's] discharge " (emphasis added)); Mt. Healthy, 429 U. S., at 286— 287. An employee who makes an unprotected statement is not immunized from discipline by the fact that this statement is surrounded by protected statements.

Nonetheless, we agree with the Court of Appeals that the District Court erred in granting summary judgment in petitioners' favor. Though Davis and Waters would have been justified in firing Churchill for the statements outlined above, there remains the question whether Churchill was actually fired because of those statements, or because of something else. See Mt. Healthy, supra, at 286-287.

Churchill has produced enough evidence to create a material issue of disputed fact about petitioners' actual motivation. Churchill had criticized the cross-training policy in the past; management had exhibited some sensitivity about the criticisms; Churchill pointed to some other conduct by hospital *682 management that, if viewed in the light most favorable to her, would show that they were hostile to her because of her criticisms. 977 F. 2d, at 1125-1126. A reasonable factfinder might therefore, on this record, conclude that petitioners actually fired Churchill not because of the disruptive things she said to Perkins-Graham, but because of nondisruptive statements about cross-training that they thought she may have made in the same conversation, or because of other statements she may have made earlier. If this is so, then the court will have to determine whether those statements were protected speech, a different matter than the one before us now.

Because of our conclusion, we need not determine whether the defendants were entitled to qualified immunity. We also need not decide whether the defendants were acting pursuant to hospital policy or custom, because that question, though argued by petitioners in their merits brief, was not presented in the petition for certiorari. See Izumi Seimitsu Kogyo Kabushiki Kaisha v. U. S. Philips Corp., 510 U. S. 27 (1993) (per curiam). Rather, we vacate the judgment of the Court of Appeals and remand the case for further proceedings consistent with this opinion.

So ordered.

Justice Souter, concurring.

I

I join Justice O'Connor's plurality opinion stating that, under the Free Speech Clause, a public employer who reasonably believes a third-party report that an employee engaged in constitutionally unprotected speech may punish the employee in reliance on that report, even if it turns out that the employee's actual remarks were constitutionally protected. I add these words to emphasize that, in order to avoid liability, the public employer must not only reasonably investigate the third-party report, but must also actually believe *683 it. Under the plurality's opinion, an objectively reasonable investigation that fails to convince the employer that the employee actually engaged in disruptive or otherwise unprotected speech does not inoculate the employer against constitutional liability. A public employer violates the Free Speech Clause, that is, by invoking a third-party report to penalize an employee when the employer, despite the report and the reasonable investigation into it, believes or genuinely suspects that the employee's speech was protected in its entirety or in that part on which the employer purports to rely in taking disciplinary action; or if the employer invokes the third-party report merely as a pretext to shield disciplinary action taken because of protected speech the employer believes or genuinely suspects that the employee uttered at another time.

First Amendment limitations on public employers, as the plurality explains, must reflect a balance of the public employer's interest in accomplishing its mission and the public employee's interest in speaking on matters of public concern. See ante, at 668-675. Where an employer penalizes an employee on the basis of a third-party report of speech that the employer should have suspected, based on the content of the report and the employer's familiarity with the employee and the workplace, to have been constitutionally protected, this balance must reflect the facts that employees' speech on matters of public concern will often (as we said of employees' union activities) "engende[r] strong emotions and giv[e] rise to active rumors," and, critically, that "the example of employees who are discharged on false charges would or might have a deterrent effect on other employees." NLRB v. Burnup & Sims, Inc., 379 U. S. 21, 23 (1964); see also Rankin v. McPherson, 483 U. S. 378, 384 (1987) ("`[T]he threat of dismissal from public employment is . . . a potent means of inhibiting speech' ") (quoting Pickering v. Board of Ed. of Township High School Dist. 205, Will Cty., 391 U. S. 563, 574 (1968)). As the plurality's opinion frankly recognizes, *684 permitting public employers to punish employees in reliance on third-party reports "involve[s] some risk of erroneously punishing protected speech." Ante, at 676.

This is a risk that the public employer's interests justify tolerating, as the plurality's opinion explains, but only when the public employer's conduct was reasonable, see ante, at 677-678, and only when the employer "really did believe" the third-party report, ante, at 679; see also ante, at 680 (an employer need not investigate further "`if the belief an employer forms supporting its adverse personnel action is "reasonable"` ") (citation omitted); ante, at 677 (courts must "look to the facts as the employer reasonably found them to be") (emphasis deleted).^[*] A public employer who did not really believe that the employee engaged in disruptive or otherwise punishable speech can assert no legitimate interest strong enough to justify chilling protected expression, whether the employer affirmatively disbelieved the third-party report or merely doubted its accuracy. Imposing liability on such an employer respects the "long-standing recognition that the First Amendment's primary aim is the full protection of speech upon issues of public concern, as well as the practical realities involved in the administration of a government office." Connick v. Myers, 461 U. S. 138, 154 (1983).

Accordingly, even though petitioners conducted an objectively reasonable investigation into Ballew's report about respondent Churchill's conversation with Perkins-Graham, I believe that petitioners' dismissal of Churchill would have violated the Free Speech Clause if after the investigation they doubted the accuracy of the report and fired Churchill for speech, or for a portion of her speech, that they genuinely suspected was nondisruptive (assuming that the speech was *685 actually on a matter of public concern). Though under the plurality's opinion the presentation of such an argument is open to Churchill on remand, Churchill would not, of course, have to rely on it if she can establish that, despite the reasonable investigation, petitioners believed that Churchill said nothing disruptive in her conversation with Perkins-Graham; that they believed that Churchill made some nondisruptive remarks to Perkins-Graham and fired her because of those remarks; or that they fired her because of nondisruptive comments about cross-training they knew she made earlier (again, assuming in each case that the speech at issue was on a matter of public concern).

II

Though Justice O'Connor's opinion speaks for just four Members of the Court, the reasonableness test it sets out is clearly the one that lower courts should apply. A majority of the Court agrees that employers whose conduct survives the plurality's reasonableness test cannot be held constitutionally liable (assuming the absence of pretext), see ante, at 679-681 (plurality opinion); post, at 686-692 (Scalia, J., concurring in judgment); and a majority (though a different one) is of the view that employers whose conduct fails the plurality's reasonableness test have violated the Free Speech Clause, see ante , at 677-678 (plurality opinion); post, at 694— 699 (Stevens, J., dissenting); see also post, at 697-698, n. 4 (Stevens, J., dissenting) ("Justice O'Connor appropriately rejects [Justice Scalia's] position, at least for those instances in which the employer unreasonably believes an incorrect report concerning speech that was in fact protected and disciplines an employee based upon that misunderstanding. I, of course, agree with Justice O'Connor that discipline in such circumstances violates the First Amendment"). Accordingly, the plurality opinion may be taken to state the holding of the Court. See Marks v. United States, 430 U. S. 188, 193-194 (1977) (discussing Book Named "John Cleland's *686 Memoirs of a Woman of Pleasure" v. Attorney General of Mass., 383 U. S. 413 (1966)).

Justice Scalia, with whom Justice Kennedy and Justice Thomas join, concurring in the judgment.

The central issue in this case is whether we shall adhere to our previously stated rule that a public employer's disciplining of an employee violates the Speech and Press Clause of the First Amendment only if it is in retaliation for the employee's speech on a matter of public concern. Justice O'Connor would add to this prohibition a requirement that the employer conduct an investigation before taking disciplinary action in certain circumstances. This recognition of a broad new First Amendment procedural right is in my view unprecedented, superfluous to the decision in the present case, unnecessary for protection of public-employee speech on matters of public concern, and unpredictable in its application and consequences.

I

I do not doubt that the First Amendment contains within it some procedural prescriptions—that in some circumstances, "the freedom of speech" recognized by the Constitution consisted of a right to speak unless and until certain procedures to prevent the speech had first been complied with. Thus, for example, I have no quarrel in principle with (though I have not inquired into the historical justification for) decisions such as Freedman v. Maryland, 380 U. S. 51 (1965), which established the administrative and judicial review provisions that a film licensing process must contain in order to avoid constituting an unconstitutional prior restraint, see Patterson v. Colorado ex rel. Attorney General of Colo., 205 U. S. 454, 462 (1907) (Holmes, J.).

We have, however, been most circumspect about acknowledging procedural components of the First Amendment. Almost all of the cases Justice O'Connor cites as exemplars are elaborations upon the limitation on defamation suits first *687 announced in New York Times Co. v. Sullivan, 376 U. S. 254 (1964). See, e. g., Bose Corp. v. Consumers Union of United States, Inc., 466 U. S. 485 (1984); Philadelphia Newspapers, Inc. v. Hepps, 475 U. S. 767 (1986); Masson v. New Yorker Magazine, Inc., 501 U. S. 496 (1991). These cases deal with alleged governmental deprivation of the freedom of speech specifically through the judicial process, in which context procedures are necessarily central to the discussion.^[*]Speiser v. Randall, 357 U. S. 513 (1958), also involved judicial (and pre-judicial adjudicative) process, holding that a state tax deduction could not be denied for a speech-related reason (advocacy of overthrow of the Government of the United States or of the State by unlawful means) by placing the burden of dis proving that speech-related reason upon the taxpayer. Moreover, although the existence of a First Amendment right was central to the Court's reasoning, the decision was squarely rested on the Due Process Clause, see id., at 529, and not on the First Amendment, see id., at 517, n. 3. The last case cited by Justice O'Connor, Freedman, supra, was, as I described earlier, a prior restraint case; review and requirement of procedures were to be expected.

In today's opinion by Justice O'Connor, our previous parsimony is abandoned, in favor of a general principle that "it is important to ensure not only that the substantive First Amendment standards are sound, but also that they are applied through reliable procedures," ante, at 669. Although we are assured that "not every procedure that may safeguard protected speech is constitutionally mandated," ante, at 670, the implication of that assurance is that many are. We never are informed how to tell mandated speechsafeguarding procedures from nonmandated ones, except for the clue that "each procedure involves a different mix of administrative burden, risk of erroneous punishment of *688 protected speech, and risk of erroneous exculpation of unprotected speech," ibid.

The proposed right to an investigation before dismissal for speech not only expands the concept of "First Amendment procedure" into brand new areas, but brings it into disharmony with our cases involving government employment decided under the Due Process Clause. As Justice O'Connor acknowledges, see ante, at 678, those cases hold that public employees who, like Churchill, lack a protected property interest in their jobs, are not entitled to any sort of a hearing before dismissal. See, e. g., Board of Regents of State Colleges v. Roth, 408 U. S. 564, 577-578 (1972). Such employees can be dismissed with impunity (insofar as federal constitutional protections are concerned) for the reason, accurate or not, that they are incompetent, that they have been guilty of unexcused absences, that they have stolen money from the faculty honor bar—or indeed for no reason at all. But under Justice O'Connor's opinion, if a reason happens to be given, and if the reason relates to speech and "there is a substantial likelihood that what was actually said was protected," (whatever that means), ante, at 677, an investigation to assure that the speech was not the sort protected by the First Amendment must be conducted—after which, presumably, the dismissal can still proceed even if the speech was not what the employer had thought it was, so long as it was not speech on an issue of public importance. In the present case, for example, if the requisite "First Amendment investigation" disclosed that Nurse Churchill had not been demeaning her superiors, but had been complaining about the perennial end-of-season slump of the Chicago Cubs, her dismissal, erroneous as it was, would have been perfectly OK.

This is a strange jurisprudence indeed. And the reason it is strange is that Justice O'Connor has in effect converted the government employer's First Amendment liability with respect to "public concern" speech from liability for *689 intentional wrong to liability for mere negligence. What she proposes is, at bottom, not new procedural protections for established First Amendment rights, but rather new First Amendment rights. Pickering v. Board of Ed. of Township High School Dist. 205, Will Cty., 391 U. S. 563 (1968), did not require government-employer "protection" of "public concern" speech, but merely forbade governmentemployer hostility to such speech. "[I]t is essential," Pickering said, "that [public employees] be able to speak out freely on such questions without fear of retaliatory dismissal." Id., at 572 (emphasis added). See also Connick v. Myers, 461 U. S. 138, 149 (1983) (same). The critical inquiry for the factfinder in these cases is whether the employment decision was, "in fact, made in retaliation for [the] exercise of the constitutional right of free speech." Perry v. Sindermann, 408 U. S. 593, 598 (1972). A category of employee speech is certainly not being "retaliated against" if it is no more and no less subject to being mistaken for a disciplinable infraction than is any other category of speech or conduct.

II

The creation of procedural First Amendment rights in this case is all the more remarkable because it is unnecessary to the disposition of the matter. After imposing the new duty upon government employers, Justice O'Connor's opinion concludes that it was satisfied anyway—i. e., that the investigation conducted by the hospital was "entirely reasonable." Ante, at 680. And then, to make the creation of the new duty doubly irrelevant, it finds that the case must be remanded anyway for a pretext inquiry: whether "petitioners actually fired Churchill not because of the disruptive things she said to Perkins-Graham, but because of nondisruptive statements about cross-training that they thought she may have made in the same conversation, or because of other statements she may have made earlier." Ante, at 682; see also ante, at 682-685 (Souter, J., concurring). Surely this *690 offends the doctrine that constitutional questions that need not be addressed should be avoided.

The requirement of a pretext inquiry, I think, renders creation of the new First Amendment right of investigation not only superfluous to the disposition of the present case, but superfluous to the protection of previously established speech rights. Justice O'Connor makes no attempt to justify the right of investigation on historical grounds (it is quite unheard of). The entire asserted basis for it is pragmatic and functional: without it the government employee's right not to be fired for his speech cannot be protected. The availability of a pretext inquiry disproves that argument. Judicial inquiry into the genuineness of a public employer's asserted permissible justification for an employment decision—be it unprotected speech, general insubordination, or laziness—is all that is necessary to avoid the targeting of "public interest" speech condemned in Pickering.

Our cases have hitherto considered this sort of inquiry all the protection needed. Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274 (1977), involved an arguably weaker case for the public employer than the present one, in that there was a "mixed motive" for the disciplinary action—that is, the employer admitted that the "public concern" speech was part of the reason for the discharge, but asserted that other valid reasons were in any event sufficient. In deciding that case, we found no need to invent procedural requirements, but simply directed the District Court "to determine whether the Board had shown by a preponderance of the evidence that it would have reached the same decision as to respondent's [e]mployment even in the absence of the protected conduct." Id. , at 287. The objective, we said, was to "protec[t] against the invasion of constitutional rights without commanding undesirable consequences not necessary to the assurance of those rights." Ibid.

The Court considers "pretext" analysis sufficient in many other areas. See, e. g., Eastman Kodak Co. v. Image Techni- *691 cal Services, Inc., 504 U. S. 451, 484 (1992) (antitrust laws); Hernandez v. New York, 500 U. S. 352, 363-364 (1991) (plurality opinion) (constitutionality of peremptory challenges); Patterson v. McLean Credit Union, 491 U. S. 164, 187-188 (1989) (employment discrimination suit under 42 U. S. C. § 1981); New York v. Burger, 482 U. S. 691, 716-717, n. 27 (1987) (Fourth Amendment challenge to administrative searches); Sure-Tan, Inc. v. NLRB, 467 U. S. 883, 895-896, n. 6 (1984) (unfair labor practice suit under the National Labor Relations Act); Geduldig v. Aiello, 417 U. S. 484, 496— 497, n. 20 (1974) (Equal Protection Clause sex-discrimination claim against legislation); McDonnell Douglas Corp. v. Green, 411 U. S. 792, 804-805 (1973) (discrimination claim under Title VII). And it considers "pretext" analysis sufficient in other First Amendment contexts. For example, in Renton v. Playtime Theatres, Inc., 475 U. S. 41, 54 (1986), after holding that zoning laws restricting the location of movie theaters do not violate the First Amendment unless they are a pretext for preventing free speech, we did not think it necessary to prescribe "reasonable" procedures for zoning commissions across the Nation; we left it to factfinders to determine whether zoning regulations are prompted by legitimate or improper factors. See also Arcara v. Cloud Books, Inc., 478 U. S. 697, 708 (1986) (O'Connor, J., concurring) (same). There is no reason why the same approach should not suffice here.

Justice Stevens believes that "pretext" review is inadequate, since "it provides less protection for a fundamental constitutional right than the law ordinarily provides for less exalted rights"; and "[o]rdinarily," he contends, "when someone acts to another person's detriment based upon a factual judgment, the actor assumes the risk that an impartial adjudicator may come to a different conclusion." Post, at 696. But that is true in contractual realms only to the extent that the contract provides a "right" whose elimination constitutes a legal "detriment." An employee dismissable at will can *692 be fired on the basis of an erroneous factual judgment, with no legal recourse—which is what happened here. Churchill also had a non contractual right: the right not to be dismissed (even from an at-will government job) in retaliation for her expression of views on a matter of public concern. That right was not violated, since she was dismissed for another reason, erroneous though it may have been. The issue before us has nothing to do with according the deprivation of a right the ordinary degree of protection; it has to do with expanding the protection accorded a government employee's public interest speech from (1) protection against retaliation, to (2) protection against retaliation and mistake.

III

The approach to this case adopted by Justice O'Connor's opinion provides more questions than answers, subjecting public employers to intolerable legal uncertainty. Despite the difficulties courts already encounter in distinguishing between protected and unprotected speech, see, e. g., Miller v. California, 413 U. S. 15, 22 (1973), and in determining whether speech pertains to a matter of public concern, compare O'Connor v. Steeves, 994 F. 2d 905, 915 (CA1), cert. denied, 510 U. S. 1024 (1993), with Gillum v. City of Kerrville, 3 F. 3d 117, 120-121 (CA5 1993), cert. denied, 510 U. S. 1072 (1994), Justice O'Connor creates yet another speechrelated puzzlement that government employers, judges, and juries must struggle to solve. The new constitutional duty to provide certain minimum procedural protections is triggered when "an employment action is based on what an employee supposedly said, and a reasonable supervisor would recognize that there is a substantial likelihood that what was actually said was protected," ante, at 677. But on what does the "reasonable supervisor" base his judgment as to whether "there is a substantial likelihood that what was actually said was protected?" Can he base it upon the report of what was said? Seemingly not, since otherwise Justice O'Connor *693 would not have found the minimum procedural protection of investigation to have been required in the present case (the report of Churchill's conversation gave no hint of protected speech). It remains entirely unclear what the employer's judgment must be based on. To avoid liability, he had better assume that it must be based on what was actually said —which means that he had better investigate the incident in order to determine whether he has an obligation to investigate the incident. Hopefully I am wrong, however, and (despite today's holding) the basis for judging whether investigation is required will be solely the report. Then the public employer will only have to figure out what a hypothetical reasonable supervisor would infer about actual speech from that report, and then determine whether that constructed "actual speech" has a substantial likelihood of being on a matter of public concern. May the employer at least assume that no investigation is required if the report does not mention speech? Or can he be liable if the recommended basis for the discipline (for example, "disrupting the workplace") had a substantial likelihood of involving speech which would have had a substantial likelihood of being on a subject of public concern? I suppose ultimately it will be up to the jury to answer all these nice, once-removed questions. Or come to think of it, perhaps it will be up to the judge. Justice O'Connor does not specify whether all this is a question of law or fact.

Justice O'Connor states that "employer decisionmaking will not be unduly burdened by having courts look to the facts as the employer reasonably found them to be." Ante, at 677 (emphasis in original). This explains the subsequent course of events when the employer's investigation has been found reasonable: The court (or the jury) decides whether, on the facts as found by the employer, the speech was on a matter of public concern, and if not, whether the employer's reliance on the report was pretextual. But what happens when the employer's investigation has been found unreasonable? *694 I presume that there has then been established a violation of the procedural component of the First Amendment—the failure to treat possibly protected speech with the requisite "amount of care"—without regard to whether the employee's speech was in fact on a matter of public concern. Justice O'Connor does not reveal what the remedy for this violation is to be. There are various possibilities: One could say that the discharge without observance of the constitutionally requisite procedures is invalid, and must be set aside unless and until those procedures are complied with. Alternatively, one could charge the employer who failed to conduct a reasonable investigation with knowledge of the protected speech that a jury later finds—producing a sort of constructive retaliatory discharge, and entitling the employee to full reinstatement and damages. Or alternatively again, the jury could be required to determine what information a reasonable investigation would have turned up, and then to decide whether it would have been permissible for the employer to fire the employee based on that information.

These are only a few of the numerous questions left unanswered by Justice O'Connor's opinion. Loose ends are the inevitable consequence of judicial invention. We will spend decades trying to improvise the limits of this new First Amendment procedure that is unmentioned in text and unformed by tradition. It seems to me clear that game is not worth the candle, given the adequacy of "pretext" analysis to protect the constitutional interest at stake.

Justice Stevens, with whom Justice Blackmun joins, dissenting.

This is a free country. Every American has the right to express an opinion on issues of public significance. In the private sector, of course, the exercise of that right may entail unpleasant consequences. Absent some contractual or statutory provision limiting its prerogatives, a private-sector *695 employer may discipline or fire employees for speaking their minds. The First Amendment, however, demands that the government respect its employees' freedom to express their opinions on issues of public importance. As long as that expression is not unduly disruptive, it simply may not provide the basis for discipline or termination. The critical issues in a case of this kind are (1) whether the speech is protected, and (2) whether it was the basis for the sanction imposed on the employee.

Applying these standards to the case before us is quite straightforward. Everyone agrees that respondent Cheryl Churchill was fired because of what she said in a conversation with co-workers during a dinner break. Given the posture in which this case comes to us, we must assume that Churchill's statements were fully protected by the First Amendment.^[1] Nevertheless, the plurality concludes that a dismissal for speech is valid as a matter of law as long as the public employer reasonably believed that the employee's speech was unprotected. See ante, at 677-678. This conclusion is erroneous because it provides less protection for a fundamental constitutional right than the law ordinarily provides for less exalted rights, including contractual and statutory rights applicable in the private sector.

If, for example, a hospital employee had a contract providing that she could retain her job for a year if she followed the employer's rules and did competent work, that employee *696 could not be fired because her supervisor reasonably but mistakenly believed she had been late to work or given a patient the wrong medicine. Ordinarily, when someone acts to another person's detriment based upon a factual judgment, the actor assumes the risk that an impartial adjudicator may come to a different conclusion.^[2] Our legal system generally delegates the determination of facts upon which important rights depend to neutral factfinders, notwithstanding the attendant risks of error and overdeterrence.

Federal constitutional rights merit at least the normal degree of protection. Doubts concerning the ability of juries to find the truth, an ability for which we usually have high regard, should be resolved in favor of, not against, the protection of First Amendment rights. See, e. g., New York Times Co. v. Sullivan, 376 U. S. 254, 279-280 (1964). Unfortunately, the plurality underestimates the importance of freedom of speech for the more than 18 million civilian employees of this country's Federal, State, and local Governments,^[3] and subordinates that freedom to an abstract interest in bureaucratic efficiency. The need for governmental efficiency that so concerns the plurality is amply protected by the substantive *697 limits on public employees' rights of expression. See generally Connick v. Myers, 461 U. S. 138 (1983); Pickering v. Board of Ed. of Township High School Dist. 205, Will Cty., 391 U. S. 563 (1968). Efficiency does not demand an additional layer of deference to employers' "reasonable" factual errors. Today's ruling will surely deter speech that would be fully protected under Pickering and Connick.

The plurality correctly points out that we have never decided whether the governing version of the facts in public employment free speech cases is "what the government employer thought was said, or . . . what the trier of fact ultimately determines to have been said." Ante, at 664.^[4] To *698 me it is clear that the latter must be controlling. The First Amendment assures public employees that they may express their views on issues of public concern without fear of discipline or termination as long as they do so in an appropriate manner and at an appropriate time and place. A violation occurs when a public employee is fired for uttering speech on a matter of public concern that is not unduly disruptive of the operations of the relevant agency. The violation does not vanish merely because the firing was based upon a reasonable mistake about what the employee said.^[5] A First Amendment claimant need not allege bad faith; the controlling question is not the regularity of the agency's investigative procedures, or the purity of its motives, but whether the employee's freedom of speech has been "abridged."

The risk that a jury may ultimately view the facts differently from even a conscientious employer is not, as the plurality would have it, a needless fetter on public employers' ability to discharge their duties. It is the normal means by which our legal system protects legal rights and encourages those in authority to act with care. Here, for example, attention to "conclusions a jury would later draw," ante, at 676, about the content of Churchill's speech might have caused petitioners to talk to Churchill about what she said before deciding to fire her. There is nothing unfair or onerous *699 about putting the risk of error on an employer in these circumstances.^[6]

Government agencies are often the site of sharp differences over a wide range of important public issues. In offices where the First Amendment commands respect for candid deliberation and individual opinion, such disagreements are both inevitable and desirable. When those who work together disagree, reports of speech are often skewed, and supervisors are apt to misconstrue even accurate reports. The plurality, observing that managers "can spend only so much of their time on any one employment decision," ante, at 680, adopts a rule that invites discipline, rather than further discussion, when such disputes arise. That rule is unwise, for deliberation within the government, like deliberation about it, is an essential part of our "profound national commitment" to the freedom of speech. Cf. New York Times , 376 U. S., at 270. A proper regard for that principle requires that, before firing a public employee for her speech, management get its facts straight.

I would affirm the judgment of the Court of Appeals.

NOTES

[*] Richard Ruda and Glen D. Nager filed a brief for the International City/County Management Association et al. as amici curiae urging reversal.

Briefs of amici curiae urging affirmance were filed for the American Nurses Association by Ronald C. Jessamy; for the National Education Association et al. by Robert H. Chanin, Jeremiah A. Collins, and Larry P. Weinberg; and for the Southern States Police Benevolent Association et al. by J. Michael McGuinness.

Charles E. Tucker, Jr., Patricia C. Benassi, and Mary Lee Leahy filed a brief for the National Employment Lawyers Association as amicus curiae.

[*] In addition, and also because of the risk of chilling protected expression, the public employer must believe that the discipline chosen is an appropriate, and not excessive, response to the employee's speech as reported. I do not understand respondents in this case to raise any claim that the discharge was pretextual in this respect.

[*] Moreover, the remedy in that context is self-evident: remand for readjudication pursuant to the proper procedures. In the present context, by contrast, the remedy is not all clear, see infra, at 693-694.

[1] On review of the Court of Appeals' reversal of a summary judgment for petitioners, we naturally accept as true the version of Churchill's statements described in her testimony and that of two supporting witnesses. See 977 F. 2d 1114, 1118-1126 (CA7 1992). According to Churchill, Thomas Koch, and Jean Welty, the dinner-break conversation concerned the merits of hospital policy, and Churchill did not direct any "personal criticism" against her supervisors. See id., at 1118-1119, 1122. According to two other witnesses, Melanie Perkins-Graham and Mary Lou Ballew, Churchill's speech was filled with "unkind and inappropriate . . . things," "negativism," and personal comment about petitioner Cynthia Waters and the hospital administration. Id., at 1118-1119.

[2] In NLRB v. Burnup & Sims, Inc., 379 U. S. 21 (1964), two employee labor organizers were fired based upon a report that they had threatened to dynamite the employer's plant if a coming representation election was unsuccessful. The National Labor Relations Board found that the employees had never made the threatening statements. Although we recognized that the employer had acted in good faith, this Court held that the discharge "plainly violated" the organizers' right under § 8 of the National Labor Relations Act. Id., at 22. "Union activity," we observed, "often engenders strong emotions and gives rise to active rumors. A protected activity acquires a precarious status if innocent employees can be discharged while engaging in it, even though the employer acts in good faith." Id., at 23. The plurality does not explain why First Amendment rights should receive any lesser protection than the statutory right at issue in Burnup & Sims.

[3] See U. S. Dept. of Commerce, Statistical Abstract of the United States 318 (113 ed. 1993) (Table No. 500) (figure from 1991).

[4] Justice Scalia would recharacterize employees' right to free speech as a more modest protection against "retaliatory" discharges, a protection that would not extend to those terminated for speech that was fully protected but incorrectly reported. The only support he cites for this restrictive theory is that three of our prior public employment speech opinions have used the word "retaliation." See ante, at 688-689 (opinion concurring in judgment) (citing Connick v. Myers, 461 U. S. 138, 149 (1983); Perry v. Sindermann, 408 U. S. 593, 598 (1972); Pickering v. Board of Ed. of Township High School Dist. 205, Will Cty., 391 U. S. 563, 572 (1968)). Our use of that word in the cases Justice Scalia cites, however, does not resolve the present question, since none of those decisions involved any factual dispute over the content of employee speech. More importantly, other passages from two of those opinions support the view that the causal connection between the employee's speech and her discharge is all the "retaliation" that must be shown. See Perry, 408 U. S., at 598 (nonrenewal of a teacher's contract "may not be predicated on his exercise of First and Fourteenth Amendment rights"); ibid. ("[A] teacher's public criticism of his superiors on matters of public concern may be constitutionally protected and may, therefore, be an impermissible basis for termination of his employment"); Pickering, 391 U. S., at 574 ("In sum, . . . a teacher's exercise of his right to speak on issues of public importance may not furnish the basis for his dismissal from public employment"). Precedent certainly does not command Justice Scalia's approach, and nothing in the First Amendment recommends a rule that makes ignorance or mistake a complete defense for a discharge based on fully protected speech. Justice O'Connor appropriately rejects that position, at least for those instances in which the employer unreasonably believes an incorrect report concerning speech that was in fact protected and disciplines an employee based upon that misunderstanding. I, of course, agree with Justice O'Connor that discipline in such circumstances violates the First Amendment.

[5] The reasonableness of the public employer's mistake would, of course, bear on whether that employer should be liable for damages. See Butz v. Economou, 438 U. S. 478, 507 (1978) ("Federal officials will not be liable for mere mistakes in judgment, whether the mistake is one of fact or one of law"). It is wrong, however, to constrict the substantive reach of a public employee's right of free speech in response to such remedial considerations. See ante, at 677 (government employers who use reasonable procedures should be free to act "without fear [of]liability ") (emphasis added).

[6] Because there is no dispute that Churchill was fired for the content of her speech, this case does not involve the problem of determining whether the public employee would have been terminated anyway for reasons unrelated to speech. See Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274 (1977).