delivered the opinion of the Court.

Petitioner was convicted by a jury on five counts of an indictment charging him with knowingly and willfully making oral threats “to take the life of or to inflict bodily harm upon the President of the United States,” in violation of 18 U. S. C. § 871 (a). The Court of Appeals affirmed, 488 F. 2d 512 (CA5 1974), and we granted certiorari to resolve an apparent conflict among the Courts of Appeals concerning the elements of the offense proscribed by §871 (a). 419 U. S. 824 (1974). After full briefing and argument, however, we find it unnecessary to reach that question, since certain circumstances of petitioner’s trial satisfy us that the conviction must be reversed.

The record reveals that the jury retired for deliberation at 3 p. m. on the second day of petitioner’s trial. Approximately two hours later, at 4:55 p. m., the jury sent a note, signed by the foreman, to the trial judge, inquiring whether the court would “accept the Verdict — 'Guilty as charged with extreme mercy of the Court.’ ” Without notifying petitioner or his counsel, the court instructed the marshal who delivered the note “to advise the jury that the Court’s answer was in the affirmative.” *37Five minutes later, at 5 p. m., the jury returned, and the record contains the following account of the acceptance of its verdict:

“THE COURT: We understand from a note you sent to the Court the verdict finds him guilty on all five counts but that you wish to recommend extreme mercy; is that correct?

“THE FOREMAN: Yes, Your Honor.

“THE COURT: Will you please poll the jury. (Whereupon the jury was polled and all jurors answered in the affirmative.)

“THE COURT: Let the verdict be entered as the judgment of the Court. Certainly the Court will take into consideration your recommendation of mercy, but before we can act upon the case, we will have the Probation Officer make a pre-sentence investigation report. We do not know whether the man has a prior criminal record or not and we will certainly take into account what you have recommended.” 2 Tr. 192-193.1

*38Generally, a recommendation of leniency made by a jury without statutory authorization does not affect the validity of the verdict and may be disregarded by the sentencing judge. See Cook v. United States, 379 F. 2d 966, 970 (CA5 1967), and cases cited. However, in Cook, the Court of Appeals held that an exception to this general rule, requiring further inquiry by the trial court, arises where the circumstances of the recommendation cast doubt upon the unqualified nature of the verdict. Assuming the validity of the exception, we need not decide whether either the factual differences between the recommendation in Cook and that in the instant case, or petitioner’s failure to request further inquiry prior to the recording of the verdict, see Fed. Rule Crim. Proc. 31 (d), would suffice to distinguish the cases for purposes of appropriate appellate relief. See 8 J. Moore, Federal Practice ¶ 31.07 (2d ed. 1975). We deal here not merely with a potential defect in the verdict.

In Fillippon v. Albion Vein Slate Co., 250 U. S. 76 (1919), the Court observed “that the orderly conduct of a trial by jury, essential to the proper protection of the right to be heard, entitles the parties who attend for the purpose to be present in person or by counsel at all proceedings from the time the jury is impaneled until it is discharged after rendering the verdict.” Id., at 81. In applying that principle, the Court held that the trial judge in a civil case had “erred in giving a supplementary instruction to the jury in the absence of the parties and without affording them an opportunity either to be present or to make timely objection to the instruction.” Ibid.

*39In Shields v. United States, 273 U. S. 583 (1927), the Court had occasion to consider the implications of the “orderly conduct of a trial by jury” in a criminal case. The trial judge had replied to a written communication from the jury, indicating its inability to agree as to the guilt or innocence of the defendant, by sending a written direction that it must find the defendant “guilty or not guilty.” The communications were not made in open court while the defendant and his counsel were present nor were they advised of them. The jury thereupon found Shields guilty of one count with a recommendation of mercy. This Court held that a previous request by counsel for Shields and the Government that the trial judge hold the jury in deliberation until they had agreed upon a verdict “did not justify exception to the rule of orderly conduct of jury trial entitling the defendant, especially in a criminal case, to be present from the time the jury is impaneled until its discharge after rendering the verdict.” Id., at 588-589.

As in Shields, the communication from the jury in this case was tantamount to a request for further instructions. However, we need not look solely to our prior decisions for guidance as to the appropriate procedure in such a situation. Federal Rule Crim. Proc. 43 guarantees to a defendant in a criminal trial the right to be present “at every stage of the trial including the impaneling of the jury and the return of the verdict.” Cases interpreting the Rule make it clear, if our decisions prior to the promulgation of the Rule left any doubt, that the jury’s message should have been answered in open court and that petitioner’s counsel should have been given an opportunity to be heard before the trial judge responded. See, e. g., United States v. Schor, 418 F. 2d 26, 29-30 (CA2 1969); United States v. Glick, 463 F. 2d 491, 493 (CA2 1972).

*40Although a violation of Rule 43 may in some circumstances be harmless error, see Fed. Rule Crim. Proc. 52 (a); United States v. Schor, supra, the nature of the information conveyed to the jury, in addition to the manner in which it was conveyed, does not permit that conclusion in this case. The trial judge should not have confined his response to the jury’s inquiry to an indication of willingness to accept a verdict with a recommendation of “extreme mercy.” At the very least, the court should have reminded the jury that the recommendation would not be binding in any way. But see United States v. Davidson, 367 F. 2d 60 (CA6 1966).2 In addition, the response should have included the admonition that the jury had no sentencing function and should reach its verdict without regard to what sentence might be imposed. See United States v. Louie Gim Hall, 245 F. 2d 338 (CA2 1957); United States v. Glick, supra, at 494. Cf. United States v. Patrick, 161 U. S. App. D. C. 231, 494 F. 2d 1150 (1974).

The fact that the jury, which had been deliberating for almost two hours without reaching a verdict, returned a verdict of “guilty with extreme mercy” within five minutes “after being told unconditionally and unequivocally that it could recommend leniency,” United States v. Glick, supra, at 495, strongly suggests that the trial judge’s response may have induced unanimity by giving members of the jury who had previously hesitated about reaching a guilty verdict the impression that the recommendation might be an acceptable compromise. We acknowledge that the comments of the trial judge *41upon receiving the verdict may be said to have put petitioner’s counsel on notice that the jury had communicated with the court, but the only indication that the court had unilaterally communicated with the jury comes from the note itself, which the court correctly ordered to be filed in the record, with a notation as to the time of receipt and the court’s response. It appears, however, that petitioner’s counsel was not aware of the court’s communication until after we granted the petition for certiorari. In such circumstances, and particularly in light of the difficult task of the factfinder in a prosecution under § 871 (a), see Watts v. United States, 394 U. S. 705 (1969), we conclude that the combined effect of the District Court’s errors was so fraught with potential prejudice as to require us to notice them notwithstanding petitioner’s failure to raise the issue in the Court of Appeals or in this Court. Silber v. United States, 370 U. S. 717 (1962); Brotherhood of Carpenters v. United States, 330 U. S. 395, 411-412 (1947). Cf. United States v. Davidson, 367 F. 2d, at 63.

The judgment of the Court of Appeals is accordingly reversed, and the case is remanded for further proceedings consistent with this opinion.

Reversed and remanded.

concurring.

George Rogers, a 34-year-old unemployed carpenter with a 10-year history of alcoholism, wandered into the coffee shop of a Holiday Inn in Shreveport, La., early one morning, behaving in a loud and obstreperous manner. He accosted several customers and waitresses, telling them, among other things, that he was Jesus Christ and that he was opposed to President Nixon’s visiting China because the Chinese had a bomb that only *42he knew about, which might be used against the people of this country. In the course of his various outbursts, Rogers announced that he was going to go to Washington to “whip Nixon’s ass,” or to “kill him in order to save the United States.”

The local police were soon called to remove Rogers from the Holiday Inn. When the arresting officer arrived, he asked Rogers whether he had threatened the President. Rogers replied that he didn’t like the idea of the President’s going to China and making friends with the Chinese, our enemies. He told the officer, “I’m going to Washington and I’m going to beat his ass off. Better yet, I will go kill him.” Rogers added that he intended to “walk” to Washington because he didn’t like cars. Rogers was not charged with any state-law crimes, but the police reported the incident to a local Secret Service agent, who subsequently had petitioner arrested on a federal warrant.

This sad set of circumstances resulted in a five-count indictment under the “threats against the President” statute, 18 U. S. C. § 871 (a). After a jury trial, petitioner was convicted under that statute and sentenced to five years’ imprisonment, to be followed by five years of supervised probation. The Court of Appeals for the Fifth Circuit affirmed petitioner’s conviction in a brief per curiam opinion, holding that the District Court had properly instructed the jury under § 871, and that the evidence against petitioner was sufficient to sustain a conviction under that statute as properly construed.

After we granted certiorari, and after the petitioner’s brief was filed here, the Solicitor General confessed error, but on a point that had not been raised either here, in the Court of Appeals, or at trial. The Court today seizes on that point to reverse the conviction, leaving unresolved the issue that we granted certiorari to consider. Al*43though I do not disagree with the Court’s treatment of the question on which it bases its reversal today, I would reach the merits and reverse petitioner’s conviction on the grounds pressed in the Court of Appeals and in the petition for certiorari.

I

The District Court and the Court of Appeals adopted what has been termed the “objective” construction of the statute. This interpretation of § 871 originated with the early case of Ragansky v. United States, 253 F. 643 (CA7 1918), and it has been adopted by a majority of the Courts of Appeals,1 even though this Court has expressed “grave doubts” as to its correctness. Watts v. United States, 394 U. S. 705, 707 (1969). As applied in Ragansky and later cases, this construction would support the conviction of anyone making a statement that would reasonably be understood as a threat, see Roy v. United States, 416 F. 2d 874, 877 (CA9 1969), as long as the defendant intended to make the statement and knew the meaning of the words used, see Ragansky v. United States, supra, at 645.

The District Court charged the jury in accordance with the “objective construction.” The jury was instructed in effect that it was not required to find that the petitioner actually intended to kill or injure the President, or even that he made a statement that he thought might be taken as a serious threat. Instead, the jury was permitted to convict on a showing merely that *44a reasonable man in petitioner’s place would have foreseen that the statements he made would be understood as indicating a serious intention to commit the act.2 In addition, the court charged that the jury could find petitioner guilty if his statements evinced “an apparent determination to carry out the threat.” 2 Tr. 177. In my view, this construction of § 871 is too broad.

In Watts, we observed that giving § 871 an expansive construction would create a substantial risk that crude, but constitutionally protected, speech might be criminalized. The petitioner there had been convicted for telling a small group at a political rally: “If they ever make me carry a rifle the first man I want to get in my sights is L. B. J.” We held that the statement, even if “willfully and knowingly” made, was not a true “threat” but merely a form of political hyperbole. Applying the statute with an eye to the danger of encroaching on constitutionally protected speech, we held that the comment in Watts fell outside the reach of the statute as a matter of law. Although the petitioner in the present case was not at a political rally or engaged in formal political discussion, the same concern counsels against permitting the statute such a broad construction that there is a substantial risk of conviction for a merely crude or .careless expression of political enmity.

II

Both the legislative history and the purposes of the statute are inconsistent with the “objective” construction of § 871 and suggest that a narrower view of the statute is proper.

*45A

The statute was enacted in 1917 without extensive discussion. Only in the House debates is there any hint of the scope that the sponsors intended for the Act. When it was suggested that the word “willfully” be removed from the bill, Representative Volstead objected, stating that in his view, “[t]he word ‘willfully’ adds an intention to threaten, and distinguishes a case [in which the defendant does not intend to convey any threat].” Without the requirement of willfulness, he said, “a person might send innocently, without any intention to convey a threat at all, an instrument to a friend that contained a threat, and he would be guilty . . . .” 53 Cong. Rec. 9378 (1916). Arguing — successfully, as it turned out — that the word “willfully” should be left in the statute, the Congressman emphasized the importance of the subjective intention to threaten:

“[I]f this statute is to be saved at all, it seems to me it must be upon the theory that the act is willful. There is not anything in the language outside of that word to convey the idea that a threat must be an intentional threat against the President. The word ‘willful’ conveys, as ordinarily used, the idea of wrongful as well as intentional. That idea ought to be preserved so as not to make innocent acts punishable.” Id., at 9379.

Representative Webb, the only other Congressman to comment about this issue on the House floor, also understood it to require specific intent. He read it at least as restrictively as did Representative Volstead:

“If you make it a mere technical offense, you do not give him much of a chance when he comes to answer before a court and jury. I do not think we ought to be too anxious to convict a man who does *46a thing thoughtlessly. I think it ought to be a willful expression of an intent to carry out a threat against the Executive . . . Id., at 9378.3

The sponsors thus rather plainly intended the bill to require a showing that the defendant appreciated the threatening nature of his statement and intended at least to convey the impression that the threat was a serious one. The danger of making § 871 a mere “technical offense” or making “innocent acts punishable” was clear to the sponsors of the Act; their concerns should continue to inform the application of the statute today.

B

The Government argues that only the objective construction of § 871 is consistent with the purposes the statute was intended to serve. In Watts, the Government notes, we identified the interests advanced by the statute as being both “protecting the safety of [the] Chief Executive and . . . allowing him to perform his duties without interference from threats of physical violence.” 394 U. S., at 707. I adhere to that statement of the purpose of the statute; I simply do not agree that the objective construction is the necessary or even the natural means of achieving that purpose.

Plainly, threats may be costly and dangerous to society in a variety of ways, even when their authors have no *47intention whatever of carrying them out. Like a threat to blow up a building, a serious threat on the President’s life is enormously disruptive and involves substantial costs to the Government. A threat made with no present intention of carrying it out may still restrict the President’s movements and require a reaction from those charged with protecting the President. Because § 871 was intended to prevent not simply attempts on the President’s life, but also the harm associated with the threat itself, I believe that the statute should be construed to proscribe all threats that the speaker intends to be interpreted as expressions of an intent to kill or injure the President. This construction requires proof that the defendant intended to make a threatening statement, and that the statement he made was in fact threatening in nature. Under the objective construction by contrast, the defendant is subject to prosecution for any statement that might reasonably be interpreted as a threat, regardless of the speaker’s intention. In essence, the objective interpretation embodies a negligence standard, charging the defendant with responsibility for the effect of his statements on his listeners. We have long been reluctant to infer that a negligence standard was intended in criminal statutes, see Morissette v. United States, 342 U. S. 246 (1952); we should be particularly wary of adopting such a standard for a statute that regulates pure speech. See Abrams v. United States, 250 U. S. 616, 626-627 (1919) (Holmes, J., dissenting).

If § 871 has any deterrent effect, that effect is likely to work only as to statements intended to convey a threat. Statements deemed threatening in nature only upon “objective” consideration will be deterred only if persons criticizing the President are careful to give a wide berth to any comment that might be construed as threatening in nature. And that degree of deterrence *48would have substantial costs in discouraging the “uninhibited, robust, and wide-open” debate that the First Amendment is intended to protect. New York Times Co. v. Sullivan, 376 U. S. 254, 270 (1964).

I would therefore interpret § 871 to require proof that the speaker intended his statement to be taken as a threat, even if he had no intention of actually carrying it out. The proof of intention would, of course, almost certainly turn on the circumstances under which the statement was made: if a call were made to the White House threatening an attempt on the President’s life within an hour, for example, the caller might well be subject to punishment under the statute, even though he was calling from Los Angeles at the time and had neither the purpose nor the means to carry out the threat. But to permit the jury to convict on no more than a showing that a reasonably prudent man would expect his hearers to take his threat seriously is to impose an unduly stringent standard in this sensitive area.

Under the narrower construction of § 871, the jury in this case might well have acquitted, concluding that it was unlikely that Rogers actually intended or expected that his listeners would take his threat as a serious one. Because I think that the District Court’s misconstruction of the statute prejudiced petitioner in this case and may continue to do mischief in future prosecutions brought under § 871, I would reverse on this ground rather than on the Solicitor General’s confession of error.