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.
* 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. 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." 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 re viewed 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 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); Wulfv. Wichita, 883 F.2d 842 (CAlO 1989); Sims v. Metropolitan Dade County, 972 F.2d 1230 (CAll 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. 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 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). 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 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). 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, 367368 (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, 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 569571. 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 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 de termined 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 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. 0., 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 conducted. It should, however, be the care that a reasonable manager would use before making an employment decision-discharge, suspension, reprimand, or whatever elseof 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). 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 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 Churchi