delivered the opinion of the Court.

An unusual metaphor in a critical review of an unusual loudspeaker system gave rise to product disparagement litigation that presents us with a procedural question of first impression: Does Rule 52(a) of the Federal Rules of Civil Procedure prescribe the standard to be applied by the Court of Appeals in its review of a District Court’s determination that a false statement was made with the kind of “actual malice” described in New York Times Co. v. Sullivan, 376 U. S. 254, 279-280 (1964)?

In the May 1970 issue of its magazine, Consumer Reports, respondent published a seven-page article evaluating the quality of numerous brands of medium-priced loudspeakers. In a boxed-off section occupying most of two pages, respondent commented on “some loudspeakers of special interest,” *488one of which was the Bose 901 — an admittedly “unique and unconventional” system that had recently been placed on the market by petitioner.1 After describing the system and some of its virtues, and after noting that a listener “could pinpoint the location of various instruments much more easily with a standard speaker than with the Bose system,” respondent’s article made the following statements:

“Worse, individual instruments heard through the Bose system seemed to grow to gigantic proportions and tended to wander about the room. For instance, a violin appeared to be 10 feet wide and a piano stretched from wall to wall. With orchestral music, such effects seemed inconsequential. But we think they might become annoying when listening to soloists.” Plaintiff’s Exhibit 2, p. 274.

After stating opinions concerning the overall sound quality, the article concluded: “We think the Bose system is so unusual that a prospective buyer must listen to it and judge it for himself. We would suggest delaying so big an investment until you were sure the system would please you after the novelty value had worn off.” Id., at 275.

Petitioner took exception to numerous statements made in the article, and when respondent refused to publish a retraction, petitioner commenced this product disparagement action in the United States District Court for the District of Massachusetts.2 After a protracted period of pretrial *489discovery, the District Court denied respondent’s motion for summary judgment, 84 F. R. D. 682 (1980), and conducted a 19-day bench trial on the issue of liability. In its lengthy, detailed opinion on the merits of the case, 508 F. Supp. 1249 (1981), the District Court ruled in respondent’s favor on most issues.3 Most significantly, the District Court ruled that the petitioner is a “public figure” as that term is defined in Gertz *490v. Robert Welch, Inc., 418 U. S. 328, 342, 345, 351-352 (1974), for purposes of this case and therefore the First Amendment, as interpreted in New York Times Co. v. Sullivan, 376 U. S., at 279-280, precludes recovery in this product disparagement action unless the petitioner proved by clear and convincing evidence that'respondent made a false disparaging statement with “actual malice.”

On three critical points, however, the District Court agreed with petitioner. First, it found that one sentence in the article contained a “false” statement of “fact” concerning the tendency of the instruments to wander.4 Based primarily on testimony by the author of the article, the District Court found that instruments heard through the speakers tended to wander “along the wall,” rather than “about the room” as reported by respondent.6 Second, it found that *491the statement was disparaging. Third, it concluded “on the basis of proof which it considers clear and convincing, that the plaintiff has sustained its burden of proving that the defendant published a false statement of material fact with the knowledge that it was false or with reckless disregard of its truth or falsity.” 508 F. Supp., at 1277.6 Judgment was entered for petitioner on the product disparagement claim.7

The United States Court of Appeals for the First Circuit reversed. 692 F. 2d 189 (1982). The court accepted the finding that the comment about wandering instruments was *492disparaging. It assumed, without deciding, that the statement was one of fact, rather than opinion, and that it was false, observing that “stemming at least in part from the uncertain nature of the statement as one of fact or opinion, it is difficult to determine with confidence whether it is true or false.” Id., at 194. After noting that petitioner did not contest the conclusion that it was a public figure, or the applicability of the New York Times standard, the Court of Appeals held that its review of the “actual malice” determination was not “limited” to the clearly-erroneous standard of Rule 52(a); instead, it stated that it “must perform a de novo review, independently examining the record to ensure that the district court has applied properly the governing constitutional law and that the plaintiff has indeed satisfied its burden of proof.” Id., at 195. It added, however, that it “[was] in no position to consider the credibility of witnesses and must leave questions of demeanor to the trier of fact.” Ibid. Based on its own review of the record, the Court of Appeals concluded:

“[W]e are unable to find clear and convincing evidence that CU published the statement that individual instruments tended to wander about the room with knowledge that it was false or with reckless disregard of whether it was false or not. The evidence presented merely shows that the words in the article may not have described precisely what the two panelists heard during the listening test. CU was guilty of using imprecise language in the article — perhaps resulting from an attempt to produce a readable article for its mass audience. Certainly this does not support an inference of actual malice.” Id., at 197.8

*493We granted certiorari to consider whether the Court of Appeals erred when it refused to apply the clearly-erroneous standard of Rule 52(a) to the District Court’s “finding” of actual malice. 461 U. S. 904 (1983).

I — (

To place the issue m focus, it is necessary to state m somewhat greater detail (a) the evidence on the “actual malice” issue; and (b) the basis for the District Court’s determination.

Evidence of Actual Malice.

At trial petitioner endeavored to prove that the key sentence embodied three distinct falsehoods about instruments heard through the Bose system: (1) that their size seemed grossly enlarged; (2) that they seemed to move; and (3) that their movement was “about the room.”

Although a great deal of the evidence concerned the first two points, the District Court found that neither was false. It concluded that the average reader would understand that the reference to enlarged instruments was intended to describe the size of the area from which the sound seemed to emanate rather than to any perception about the actual size of the musical instruments being played, rejecting as “absurd” the notion that readers would interpret the figurative language literally. 508 F. Supp., at 1266.9 After referring to testimony explaining that “a certain degree of movement *494of the location of the apparent sound source is to be expected with all stereo loudspeaker systems,” the District Court recognized that the statement was accurate insofar as it reported that “instruments . . . tended to wander . . . .” Id., at 1267. Thus, neither the reference to the apparent size of the instruments, nor the reference to the fact that instruments appeared to move, was false.10

The statement that instruments tended to wander “about the room” was found false because what the listeners in the test actually perceived was an apparent movement back and forth along the wall in front of them and between the two speakers. Because an apparent movement “about the room” — rather than back and forth — would be so different from what the average listener has learned to expect, the District Court concluded that “the location of the movement of the apparent sound source is just as critical to a reader as the fact that movement occurred.” Ibid.

The evidence concerning respondent’s knowledge of this falsity focused on Arnold Seligson, an engineer employed by respondent. Seligson supervised the test of the Bose 901 and prepared the written report upon which the published article was based. His initial in-house report contained this sentence: ‘“Instruments not only could not be placed with precision but appeared to suffer from giganticism and a tendency to wander around the room; a violin seemed about 10 ft. wide, a piano stretched from wall to wall, etc.’ ” Id., at 1264, n. 28. Since the editorial revision from “around the room” to “about the room” did not change the meaning of the false statement, and since there was no evidence that the editors were aware of the inaccuracy in the original report, the actual-malice determination rests entirely on an evaluation of Seligson’s state of mind when he wrote his initial report, or when he checked the article against that report.

*495Seligson was deposed before trial and testified for almost six days at the trial itself. At one point in his direct examination, he responded at length to technical testimony by Dr. Bose, explaining the scientific explanation for the apparent movement of the source of sound back and forth across a wall. App. 117-122. The trial judge then questioned Selig-son, and that questioning revealed that the movement which Seligson had heard during the tests was confined to the wall.11 During his cross-examination, at counsel’s request he drew a rough sketch of the movement of the sound source that he intended to describe with the words “tended to wander about the room”; that sketch revealed a back and forth movement along the wall between the speakers. He was then asked:

“Q. Mr. Seligson, why did you use the words ‘tended to wander about the room’ to describe what you have drawn on the board?

“A. Well, I don’t know what made me pick that particular choice of words. Would you have been more sat*496isfied if we said ‘across/ — I think not — instead of before. I have the feeling you would have objected in either event. The word ‘about’ meant just as I drew it on the board. Now, I so testified in my deposition.” Id., at 169.12

The District Court’s Actual-Malice Determination.

The District Court’s reasons for finding falsity in the description of the location of the movement of the wandering instruments provided the background for its ruling on actual malice. The court concluded that “no reasonable reader” would understand the sentence as describing lateral movement along the wall. Because the “average reader” would interpret the word “about” according to its “plain ordinary meaning,” the District Court unequivocally rejected Selig-son’s testimony — and respondent’s argument — that the sentence, when read in context, could be understood to refer to lateral movement.13

*497On similar reasoning the District Court found Seligson’s above-quoted explanation of the intended meaning of the sentence incredible. The District Court reasoned:

“Thus, according to Seligson, the words used in the Article — ‘About the room’ — mean something different to him than they do to the populace in general. If Seligson is to be believed, at the time of publication of the Article he interpreted, and he still interprets today, the words ‘about the room’ to mean ‘along the wall.’ After careful consideration of Seligson’s testimony and of his demeanor at trial, the Court finds that Seligson’s testimony on this point is not credible. Seligson is an intelligent person whose knowledge of the English language cannot be questioned. It is simply impossible for the Court to believe that he interprets a commonplace word such as ‘about’ to mean anything other than its plain ordinary meaning.

“Based on the above finding that Seligson’s testimony to the contrary is not credible, the Court further finds that at the time of the Article’s publication Seligson knew that the words ‘individual instruments . . . tended to wander about the room’ did not accurately describe the effects that he and Lefkow had heard during the ‘special listening test.’ Consequently, the Court concludes, on the basis of proof which it considers clear and convincing, that the plaintiff has sustained its burden of proving that the defendant published a false statement of material fact with the knowledge that it was false or with reckless disregard of its truth or falsity.” 508 F. Supp., at 1276-1277.

*498Notably, the District Court’s ultimate determination of actual malice was framed as a conclusion and was stated in the disjunctive. Even though the District Court found it impossible to believe that Seligson — at the time of trial — was truthfully maintaining that the words “about the room” could fairly be read, in context, to describe lateral movement rather than irregular movement throughout the room, the District Court did not identify any independent evidence that Seligson realized the inaccuracy of the statement, or entertained serious doubts about its truthfulness, at the time of publication.14

II

This is a case in which two well-settled and respected rules of law point in opposite directions.

Petitioner correctly reminds us that Rule 52(a) provides:

“Findings of fact shall not be set aside unless clearly erroneous, and due regard shall be given to the opportunity of the trial court to judge of the credibility of the witnesses.”

We have repeatedly held that the Rule means what it says. Inwood Laboratories, Inc. v. Ives Laboratories, Inc., 456 U. S. 844, 855-856 (1982); Pullman-Standard v. Swint, 456 U. S. 273, 287 (1982); United States v. United States Gypsum Co., 333 U. S. 364, 394-396 (1948). It surely does not stretch the language of the Rule to characterize an inquiry into what a person knew at a given point in time as a question of “fact.”15 In this case, since the trial judge expressly commented on Seligson’s credibility, petitioner argues that the *499Court of Appeals plainly erred when it refused to uphold the District Court’s actual-malice “finding” under the clearly-erroneous standard of Rule 52(a).

On the other hand, respondent correctly reminds us that in cases raising First Amendment issues we have repeatedly held that an appellate court has an obligation to “make an independent examination of the whole record” in order to make sure that “the judgment does not constitute a forbidden intrusion on the field of free expression.” New York Times Co. v. Sullivan, 376 U. S., at 284-286. See also NAACP v. Claiborne Hardware Co., 458 U. S. 886, 933-934 (1982); Greenbelt Cooperative Publishing Assn. v. Bresler, 398 U. S. 6, 11 (1970); St. Amant v. Thompson, 390 U. S. 727, 732-733 (1968). Although such statements have been made most frequently in cases to which Rule 52(a) does not apply because they arose in state courts, respondent argues that the constitutional principle is equally applicable to federal litigation. We quite agree; surely it would pervert the concept of federalism for this Court to lay claim to a broader power of review over state-court judgments than it exercises in reviewing the judgments of intermediate federal courts.

Our standard of review must be faithful to both Rule 52(a) and the rule of independent review applied in New York Times Co. v. Sullivan. The conflict between the two rules is in some respects more apparent than real. The New York Times rule emphasizes the need for an appellate court to make an independent examination of the entire record; Rule 52(a) never forbids such an examination, and indeed our seminal decision on the Rule expressly contemplated a review of the entire record, stating that a “finding is ‘clearly erroneous’ when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed.” United States v. United States Gypsum Co., supra, at 395 (emphasis supplied). Moreover, Rule 52(a) commands that “due regard” shall be given to the trial judge’s opportunity to *500observe the demeanor of the witnesses; the constitutionally based rule of independent review permits this opportunity to be given its due. Indeed, as we previously observed, the Court of Appeals in this case expressly declined to second-guess the District Judge on the credibility of the witnesses.

The requirement that special deference be given to a trial judge’s credibility determinations is itself a recognition of the broader proposition that the presumption of correctness that attaches to factual findings is stronger in some cases than in others. The same “clearly erroneous” standard applies to findings based on documentary evidence as to those based entirely on oral testimony, see United States Gypsum Co., supra, at 394, but the presumption has lesser force in the former situation than in the latter. Similarly, the standard does not change as the trial becomes longer and more complex, but the likelihood that the appellate court will rely on the presumption tends to increase when trial judges have lived with the controversy for weeks or months instead of just a few hours.16 One might therefore assume that the cases in which the appellate courts have a duty to exercise *501independent review are merely those in which the presumption that the trial court’s ruling is correct is particularly weak. The difference between the two rules, however, is much more than a mere matter of degree. For the rule of independent review assigns to judges a constitutional responsibility that cannot be delegated to the trier of fact, whether the factfinding function be performed in the particular case by a jury or by a trial judge.

Rule 52(a) applies to findings of fact, including those described as “ultimate facts” because they may determine the outcome of litigation. See Pullman-Standard v. Swint, 456 U. S., at 287. But Rule 52(a) does not inhibit an appellate court’s power to correct errors of law, including those that may infect a so-called mixed finding of law and fact, or a finding of fact that is predicated on a misunderstanding of the governing rule of law. See ibid.; Inwood Laboratories, Inc. v. Ives Laboratories, Inc., 456 U. S., at 855, n. 15. Nor does Rule 52(a) “furnish particular guidance with respect to distinguishing law from fact.” Pullman Standard v. Swint, 456 U. S., at 288. What we have characterized as “the vexing nature” of that distinction, ibid., does not, however, diminish its importance, or the importance of the principles that require the distinction to be drawn in certain cases.17

In a consideration of the possible application of the distinction to the issue of “actual malice,” at least three characteristics of the rule enunciated in the New York Times case are *502relevant. First, the common-law heritage of the rule itself assigns an especially broad role to the judge in applying it to specific factual situations. Second, the content of the rule is not revealed simply by its literal text, but rather is given meaning through the evolutionary process of common-law adjudication; though the source of the rule is found in the Constitution, it is nevertheless largely a judge-made rule of law. Finally, the constitutional values protected by the rule make it imperative that judges — and in some cases judges of this Court — make sure that it is correctly applied. A few words about each of these aspects of the rule are appropriate.

The federal rule that prohibits a public official from recovering damages for a defamatory falsehood unless he proves that the false “statement was made with ‘actual malice’ — that is, with knowledge that it was false or with reckless disregard of whether it was false or not,” New York Times, 376 U. S., at 279-280, has its counterpart in rules previously adopted by a number of state courts and extensively reviewed by scholars for generations.18 The earlier defamation cases, in turn, have a kinship to English cases considering the kind of motivation that must be proved to support a common-law action for deceit.19 It has long been recognized that the formulation of a rule of this kind “allows the judge the maximum of power in passing judgment in the particular case.”20 *503Moreover, the exercise of this power is the process through which the rule itself evolves and its integrity is maintained.21 As we have explained, the meaning of some concepts cannot be adequately expressed in a simple statement:

“These considerations fall short of proving St. Amant’s reckless disregard for the accuracy of his statements about Thompson. ‘Reckless disregard,’ it is true, cannot be fully encompassed in one infallible definition. Inevitably its outer limits will be marked out through case-by-case adjudication, as is true with so many legal standards for judging concrete cases, whether the standard is provided by the Constitution, statutes, or case law. Our cases, however, have furnished meaningful guidance for the further definition of a reckless publication.” St. Amant v. Thompson, 390 U. S., at 730-731.

When the standard governing the decision of a particular case is provided by the Constitution, this Court’s role in marking out the limits of the standard through the process of case-by-case adjudication is of special importance. This process has been vitally important in cases involving restrictions on the freedom of speech protected by the First Amendment, particularly in those cases in which it is contended that the communication in issue is within one of the few classes of “unprotected” speech.

The First Amendment presupposes that the freedom to speak one’s mind is not only an aspect of individual liberty— and thus a good unto itself — but also is essential, to the common quest for truth and the vitality of society as a *504whole. Under our Constitution “there is no such thing as a false idea. However pernicious an opinion may seem, we depend for its correction not on the conscience of judges and juries but on the competition of other ideas.” Gertz v. Robert Welch, Inc., 418 U. S., at 339-340 (footnote omitted). Nevertheless, there are categories of communication and certain special utterances to which the majestic protection of the First Amendment does not extend because they “are no essential part of any exposition of ideas, and are of such slight social value as a step to truth that any benefit that may be derived from them is clearly outweighed by the social interest in order and morality.” Chaplinsky v. New Hampshire, 315 U. S. 568, 572 (1942).

Libelous speech has been held to constitute one such category, see Beauharnais v. Illinois, 343 U. S. 250 (1952); others that have been held to be outside the scope of the freedom of speech are fighting words, Chaplinsky v. New Hampshire, supra, incitement to riot, Brandenburg v. Ohio, 395 U. S. 444 (1969), obscenity, Roth v. United States, 354 U. S. 476 (1957), and child pornography, New York v. Ferber, 458 U. S. 747 (1982).22 In each of these *505areas, the limits of the unprotected category, as well as the unprotected character of particular communications, have been determined by the judicial evaluation of special facts that have been deemed to have constitutional significance. In such cases, the Court has regularly conducted an independent review of the record both to be sure that the speech in question actually falls within the unprotected category and to confine the perimeters of any unprotected category within acceptably narrow limits in an effort to ensure that protected expression will not be inhibited. Providing triers of fact with a general description of the type of communication whose content is unworthy of protection has not, in and of itself, served sufficiently to narrow the category, nor served to eliminate the danger that decisions by triers of fact may inhibit the expression of protected ideas.23 The principle of viewpoint neutrality that underlies the First Amendment itself, see Police Department of Chicago v. Mosley, 408 U. S. 92, 95-96 (1972), also imposes a special responsibility on judges whenever it is claimed that a particular communication is unprotected. See generally Terminiello v. Chicago, 337 U. S. 1, 4 (1949).

We have exercised independent judgment on the question whether particular remarks “were so inherently inflammatory as to come within that small class of 'fighting words’ which are ‘likely to provoke the average person to retaliation, and thereby cause a breach of the peace,’” Street v. New York, 394 U. S. 576, 592 (1969), and on the analogous question whether advocacy is directed to inciting or producing *506imminent lawless action, Hess v. Indiana, 414 U. S. 105, 108-109 (1973) (per curiam); compare id., at 111 (Rehnquist, J., dissenting) (“The simple explanation for the result in this case is that the majority has interpreted the evidence differently from the courts below”); Edwards v. South Carolina, 372 U. S. 229, 235 (1963) (recognizing duty “to make an independent examination of the whole record”); Pennekamp v. Florida, 328 U. S. 331, 335 (1946) (“[W]e are compelled to examine for ourselves the statements in issue ... to see whether or not they do carry a threat of clear and present danger ... or whether they are of a character which the principles of the First Amendment. . . protect”).24

Similarly, although under Miller v. California, 413 U. S. 15 (1973), the questions of what appeals to “prurient interest” and what is “patently offensive” under the community standard obscenity test are “essentially questions of fact,” id., at 30, we expressly recognized the “ultimate power of appellate courts to conduct an independent review of constitutional claims when necessary,” id., at 25.25 We have therefore *507rejected the contention that a jury finding of obscenity ml non is insulated from review so long as the jury was properly instructed and there is some evidence to support its findings, holding that substantive constitutional limitations govern. In Jenkins v. Georgia, 418 U. S. 153, 159-161 (1974), based on an independent examination of the evidence — the exhibition of a motion picture — the Court held that the film in question “could not, as a matter of constitutional law, be found to depict sexual conduct in a patently offensive way . . . .” Id., at 161.26 And in its recent opinion identifying a new category of unprotected expression — child pornography — the Court expressly anticipated that an “independent *508examination” of the allegedly unprotected material may be necessary “to assure ourselves that the judgment. . . ‘does not constitute a forbidden intrusion on the field of free expression.’” New York v. Ferber, 458 U. S., at 774, n. 28 (quoting New York Times Co. v. Sullivan, 376 U. S., at 285).

Hence, in New York Times Co. v. Sullivan, after announcing the constitutional requirement for a finding of “actual malice” in certain types of defamation actions, it was only natural that we should conduct an independent review of the evidence on the dispositive constitutional issue. We explained our action as follows:

“This Court’s duty is not limited to the elaboration of constitutional principles; we must also in proper cases review the evidence to make certain that those principles have been constitutionally applied. This is such a case, particularly since the question is one of alleged trespass across ‘the line between speech unconditionally guaranteed and speech which may legitimately be regulated.’ Speiser v. Randall, 357 U. S. 513, 525. In cases where that line must be drawn, the rule is that we ‘examine for ourselves the statements in issue and the circumstances under which they were made to see . . . whether they are of a character which the principles of the First Amendment, as adopted by the Due Process Clause of the Fourteenth Amendment, protect.’ Pennekamp v. Florida, 328 U. S. 331, 335; see also One, Inc. v. Olesen, 355 U. S. 371; Sunshine Book Co. v. Summerfield, 355 U. S. 372. We must ‘make an independent examination of the whole record,’ Edwards v. South Carolina, 372 U. S. 229, 235, so as to assure ourselves that the judgment does not constitute a forbidden instrusion on the field of free expression.” 376 U. S., at 285 (footnote omitted).27

*509In Time, Inc. v. Pape, 401 U. S. 279 (1971), a case in which the Federal District Court had entered a directed verdict, we again conducted an independent examination of the evidence on the question of actual malice, labeling our definition of “actual malice” as a “constitutional rule” and stating that the question before us was whether that rule had been correctly applied to the facts of the case, id., at 284. Again we stated that independent inquiries “of this kind are familiar under the settled principle that ‘[i]n cases in which there is a claim of denial of rights under the Federal Constitution, this Court is not bound by the conclusions of lower courts, but will re*510examine the evidentiary basis on which those conclusions are founded,’” noting that “in cases involving the area of tension between the First and Fourteenth Amendments on the one hand and state defamation laws on the other, we have frequently had occasion to review ‘the evidence in the . . . record to determine whether it could constitutionally support a judgment’ for the plaintiff.” Ibid, (citations omitted).28

In Monitor Patriot Co. v. Roy, 401 U. S. 265, 277 (1971) the Court held “as a matter of constitutional law” that the jury could not be allowed to determine the relevance of a defamatory statement to the plaintiff’s status as a public figure. We explained that the jury’s application of such a standard “is unlikely to be neutral with respect to the content of speech and holds a real danger of becoming an instrument for the suppression of those ‘vehement, caustic, and sometimes unpleasantly sharp attacks,’ New York Times, supra, at 270, which must be protected if the guarantees of the First and Fourteenth Amendments are to prevail.” Ibid.29

The requirement of independent appellate review reiterated in New York Times Co. v. Sullivan is a rule of federal constitutional law. It emerged from the exigency of deciding concrete cases; it is law in its purest form under our common-law heritage. It reflects a deeply held conviction that judges — and particularly Members of this Court — must *511exercise such review in order to preserve the precious liberties established and ordained by the Constitution. The question whether the evidence in the record in a defamation case is of the convincing clarity required to strip the utterance of First Amendment protection is not merely a question for the trier of fact. Judges, as expositors of the Constitution, must independently decide whether the evidence in the record is sufficient to cross the constitutional threshold that bars the entry of any judgment that is not supported by clear and convincing proof of “actual malice.”

r — 4 hH

The Court of Appeals was correct in its conclusions (1) that there is a significant difference between proof of actual malice30 and mere proof of falsity, and (2) that such additional proof is lacking in this case.

The factual portion of the District Court’s opinion may fairly be read as including the following findings: (1) Selig-son’s actual perception of the apparent movement of the sound source at the time the Bose 901 was tested was “along the wall” rather than “about the room”; (2) even when the words in the disputed sentence are read in the context of the entire article, neither the “average reader,” nor any other intelligent person, would interpret the word “about” to mean “across”; (3) Seligson is an intelligent, well-educated person; (4) the words “about the room” have the same meaning for Seligson as they do for the populace in general; and (5) although he was otherwise a credible witness, Seligson’s testimony that (a) he did not “know what made me pick that *512particular choice of words” and (b) that the word “about” meant what he had drawn on the board, is not credible.

When the testimony of a witness is not believed, the trier of fact may simply disregard it. Normally the discredited testimony is not considered a sufficient basis for drawing a contrary conclusion. See Moore v. Chesapeake & Ohio R. Co., 340 U. S. 573, 575 (1951). In this case the trial judge found it impossible to believe that Seligson continued to maintain that the word “about” meant “across.” Seligson’s testimony does not rebut any inference of actual malice that the record otherwise supports, but it is equally clear that it does not constitute clear and convincing evidence of actual malice. Seligson displayed a capacity for rationalization. He had made a mistake and when confronted with it, he refused to admit it and steadfastly attempted to maintain that no mistake had been made — that the inaccurate was accurate. That attempt failed, but the fact that he made the attempt does not establish that he realized the inaccuracy at the time of publication.

Aside from Seligson’s vain attempt to defend his statement as a precise description of the nature of the sound movement, the only evidence of actual malice on which the District Court relied was the fact that the statement was an inaccurate description of what Seligson had actually perceived. Seligson of course had insisted “I know what I heard.” The trial court took him at his word, and reasoned that since he did know what he had heard, and he knew that the meaning of the language employed did not accurately reflect what he heard, he must have realized the statement was inaccurate at the time he wrote it. “Analysis of this kind may be adequate when the alleged libel purports to be an eyewitness or other direct account of events that speak for themselves.” Time, Inc. v. Pape, 401 U. S., at 285. See generally The Santissima Trinidad, 7 Wheat. 283, 338-339 (1822). Here, however, adoption of the language chosen was “one of a number of possible rational interpretations” of an event “that bristled with ambiguities” and descriptive challenges for the writer. *513Time, Inc. v. Pape, supra, at 290. The choice of such language, though reflecting a misconception, does not place the speech beyond the outer limits of the First Amendment’s broad protective umbrella. Under the District Court’s analysis, any individual using a malapropism might be liable, simply because an intelligent speaker would have to know that the term was inaccurate in context, even though he did not realize his folly at the time.

The statement in this case represents the sort of inaccuracy that is commonplace in the forum of robust debate to which the New York Times rule applies. 401 U. S., at 292. “Realistically, . . . some error is inevitable; and the difficulties of separating fact from fiction convinced the Court in New York Times, Butts, Gertz, and similar cases to limit liability to instances where some degree of culpability is present in order to eliminate the risk of undue self-censorship and the suppression of truthful material.” Herbert v. Lando, 441 U. S. 153, 171-172 (1979). “[E]rroneous statement is inevitable in free debate, and . . . must be protected if the freedoms of expression are to have the ‘breathing space’ that they ‘need ... to survive.’” New York Times Co. v. Sullivan, 376 U. S., at 271-272 (citation omitted).

The Court of Appeals entertained some doubt concerning the ruling that the New York Times rule should be applied to a claim of product disparagement based on a critical review of a loudspeaker system. We express no view on that ruling, but having accepted it for purposes of deciding this case, we agree with the Court of Appeals that the difference between hearing violin sounds move around the room and hearing them wander back and forth fits easily within the breathing space that gives life to the First Amendment. We may accept all of the purely factual findings of the District Court and nevertheless hold as a matter of law that the record does not contain clear and convincing evidence that Seligson or his employer prepared the loudspeaker article with knowledge that it contained a false statement, or with reckless disregard of the truth.

*514It may well be that in this case, the “finding” of the District Court on the actual-malice question could have been set aside under the clearly-erroneous standard of review, and we share the concern of the Court of Appeals that the statements at issue tread the line between fact and opinion. Moreover, the analysis of the central legal question before us may seem out of place in a case involving a dispute about the sound quality of a loudspeaker. But though the question presented reaches us on a somewhat peculiar wavelength, we reaffirm the principle of independent appellate review that we have applied uncounted times before. We hold that the clearly-erroneous standard of Rule 52(a) of the Federal Rules of Civil Procedure does not prescribe the standard of review to be applied in reviewing a determination of actual malice in a case governed by New York Times Co. v. Sullivan.31 Appellate judges in such a case must exercise independent judgment and determine whether the record establishes actual malice with convincing clarity.

The judgment of the Court of Appeals is affirmed.

It is so ordered.

dissenting.

Although I do not believe that the “reckless disregard” component of the New York Times malice standard is a question of historical fact, I agree with Justice Rehnquist that the actual-knowledge component surely is. Here, the District Court found that the defamatory statement was written with actual knowledge of falsity. The Court of Appeals thus erred in basing its disagreement with the District Court on its de novo review of the record. The majority is today equally in error. I would remand to the Court of Appeals so that it may perform its task under the proper standard.

with whom Justice O’Connor joins, dissenting.

There is more than one irony in this “Case of the Wandering Instruments,” which subject matter makes it sound more like a candidate for inclusion in the “Adventures of Sherlock Holmes” than in a casebook on constitutional law. It is ironic in the first place that a constitutional principle which originated in New York Times Co. v. Sullivan, 376 U. S. 254 (1964), because of the need for freedom to criticize the conduct of public officials is applied here to a magazine’s false statements about a commercial loudspeaker system.

It is also ironic that, in the interest of protecting the First Amendment, the Court rejects the “clearly erroneous” standard of review mandated by Federal Rule of Civil Procedure 52(a) in favor of a “de novo” standard of review for the “constitutional facts” surrounding the “actual malice” determination. But the facts dispositive of that determination — actual knowledge or subjective reckless disregard for truth — involve no more than findings about the mens rea of an author, findings which appellate courts are simply ill-prepared to make in any context, including the First Amendment context. Unless “actual malice” now means something different from the definition given to the term 20 years ago by this *516Court in New York Times, I do not think that the constitutional requirement of “actual malice” properly can bring into play any standard of factual review other than the “clearly erroneous” standard.

In this case the District Court concluded by what it found to be clear and convincing evidence that respondent’s engineer Arnold Seligson had written the defamatory statement about Bose’s product with actual knowledge that it was false. It reached that conclusion expressly relying on its determination about the credibility of Seligson’s testimony. 508 F. Supp. 1249, 1276-1277 (1981). On appeal there was no issue as to whether the District Court had properly understood what findings were legally sufficient to establish “actual malice” nor was there any issue as to the necessary quantum of proof nor the proper allocation of the burden of proof of “actual malice.” The issue on appeal thus was only the propriety of the District Court’s factual conclusion that Bose had actually proved “actual malice” in this case. Yet the Court of Appeals never rebutted the District Court’s conclusion that Seligson had actual knowledge that what he printed was false. Instead it concluded after de novo review that Seligsoh’s language was merely “imprecise” and that as such, it would not “support an inference of actual malice.” 692 F. 2d 189, 197 (1982).

It is unclear to me just what that determination by the Court of Appeals has to do with the mens rea conclusion necessary to the finding of “actual malice” and with the District Court’s finding of actual knowledge here. In approving the Court of Appeals’ de novo judgment on the “actual malice” question, for all the factual detail and rehearsal of testimony with which the majority’s opinion is adorned, the Court never quite comes to grips with what factual finding it must focus on. At one point we are told that “[t]he statement in this case represents the sort of inaccuracy that is commonplace in the forum of robust debate to which the New York Times rule applies,” ante, at 513, suggesting that the disparaging *517statement was perhaps not even false, or at any rate not false enough. One paragraph later, we are told that “as a matter of law . . . the record does not contain clear and convincing evidence that Seligson or his employer prepared the loudspeaker article with knowledge that it contained a false statement, or with reckless disregard of the truth.” Ante, at 513. The Court remarks that the question presented “reaches us on a somewhat peculiar wavelength,” ante, at 514, but that is scarcely a reason for transmitting the answer on an equally peculiar wavelength.

In my view the problem results from the Court’s attempt to treat what is here, and in other contexts always has been, a pure question of fact, as something more than a fact — a so-called “constitutional fact.” The Court correctly points out that independent appellate review of facts underlying constitutional claims has been sanctioned by previous decisions of this Court where “a conclusion of law as to a Federal right and a finding of fact are so intermingled as to make it necessary, in order to pass upon the Federal question, to analyze the facts.” Fiske v. Kansas, 274 U. S. 380, 385-386 (1927). But in other contexts we have always felt perfectly at ease leaving state-of-mind determinations, such as the actual knowledge and recklessness determinations involved here, to triers of fact with only deferential appellate review— for example, in criminal cases where the burden of proving those facts is even greater than the “clear and convincing” standard applicable under New York Times.1

*518Presumably any doctrine of “independent review” of facts exists, not so that an appellate court may inexorably place its thumb on the scales in favor of the party claiming the constitutional right, but so that perceived shortcomings of the trier of fact by way of bias or some other factor may be compensated for.2 But to me, the only shortcoming here is an *519appellate court’s inability to make the determination which the Court mandates today — the de novo determination about the state of mind of a particular author at a particular time. Although there well may be cases where the “actual malice” determination can be made on the basis of objectively reviewable facts in the record, it seems to me that just as often it is made, as here, on the basis of an evaluation of the credibility of the testimony of the author of the defamatory statement. I am at a loss to see how appellate courts can even begin to make such determinations. In any event, surely such determinations are best left to the trial judge.

It is of course true as the Court recognizes that “where particular speech falls close to the line separating the lawful and the unlawful, the possibility of mistaken factfinding— inherent in all litigation — will create the danger that the legitimate utterance will be penalized.” Speiser v. Randall, 357 U. S. 513, 526 (1958). But the New York Times rule adequately addresses the need to shield protected speech from the risk of erroneous factfinding by placing the burden of proving “actual malice” on the party seeking to penalize expression. I agree with Justice Harlan who, in commenting on the inappropriateness of de novo fact review of the “actual malice” determination, concluded that he could not

“discern in those First Amendment considerations that led us to restrict the States’ powers to regulate defamation of public officials any additional interest that is not served by the actual-malice rule of New York Times, supra, but is substantially promoted by utilizing [an appellate court] as the ultimate arbiter of factual disputes in those libel cases where no unusual factors, such as allegations of harassment or the existence of a jury verdict resting on erroneous instructions . . . are present.” Time, Inc. v. Pape, 401 U. S. 279, 294 (1971) (dissenting opinion).

I think that the issues of “falsity” and “actual malice” in this case may be close questions, but I am convinced, that the *520District Court, which heard the principal witness for the respondent testify for almost six days during the trial, fully understood both the applicable law and its role as a finder of fact. Because it is not clear to me that the de novo findings of appellate courts, with only bare records before them, are likely to be any more reliable than the findings reached by trial judges, I cannot join the majority’s sanctioning of factual second-guessing by appellate courts. I believe that the primary result of the Court’s holding today will not be greater protection for First Amendment values, but rather only lessened confidence in the judgments of lower courts and more entirely factbound appeals.

I continue to adhere to the view expressed in Pullman-Standard v. Swint, 456 U. S. 273, 287 (1982), that Rule 52(a) “does not make exceptions or purport to exclude certain categories of factual findings from the obligation of a court of appeals to accept a district court’s findings unless clearly erroneous.” There is no reason to depart from that rule here, and I would therefore reverse and remand this case to the Court of Appeals so that it may apply the “clearly erroneous” standard of review to the factual findings of the District Court.