delivered the opinion of the Court.

Respondent was tried and convicted of robbery in the District Court for the District of Columbia.^[1] During cross-examination at trial the prosecutor asked respondent why he had not given the police his alibi when he was questioned shortly after his arrest. The trial court instructed the jury to disregard the colloquy but refused[*173] to declare a mistrial. The Court of Appeals for the District of Columbia Circuit reversed, holding that inquiry into respondent’s prior silence impermissibly prejudiced his defense and infringed upon his right to remain silent under Miranda v. Arizona, 384 U. S. 436, 468 n. 37 (1966). We granted certiorari, 419 U. S. 1045, because of a conflict among the Courts of Appeals over whether a defendant can be cross-examined about his silence during police interrogation,^[2] and because of the importance of this question to the administration of justice.

We find that the probative value of respondent’s pretrial silence in this case was outweighed by the prejudicial impact of admitting it into evidence. Affirming the judgment on this ground, we have no occasion to reach the broader constitutional question that supplied an alternative basis for the decision below.

I

On June 1, 1971, Lonnie Arrington reported to police that he had been attacked and robbed by a group of five men. Initially, he claimed that $65 had been stolen, but he later changed the amount to $96 after consulting with his wife. As the police were preparing to accompany Arrington through the neighborhood in search of the attackers, he observed two men and identified one of them as one of his assailants. When the police gave chase, the two men fled but one was immediately cap[*174] tured. The victim identified respondent Hale as one of the robbers.

Respondent was then arrested, taken to the police station, and advised of his right to remain silent. He was searched and found to be in possession of $158 in cash. An officer then asked: “Where did you get the money?” Hale made no response.

At trial respondent took the witness stand in his own defense. He acknowledged having met Arrington in a shoe store on the day in question. Hale stated that, after the meeting, he was approached by three men who inquired whether Arrington had any money, to which Hale replied he “didn’t know.” From there respondent claimed he went to a narcotics treatment center, where he remained until after the time of the robbery. According to his testimony he left the center with a friend who subsequently purchased narcotics. Shortly after the transaction they were approached by the police. Hale testified that he fled because he feared being found in the presence of a person carrying narcotics. He also insisted that his estranged wife had received her welfare check on that day and had given him approximately $150 to purchase some money orders for her as he had done on several prior occasions.

In an effort to impeach Hale’s explanation of his possession of the money, the prosecutor caused Hale to admit on cross-examination that he had not offered the exculpatory information to the police at the time of his arrest:

“Q. Did you in any way indicate [to the police] where that money came from?

“A. No, I didn’t.

“Q. Why not?

“A. I didn’t feel that it was necessary at the time.”

[*175] The Government takes the position that since the respondent chose to testify in his own behalf, it was permissible to impeach his credibility by proving that he had chosen to remain silent at the time of his arrest.^[3] For this proposition the Government relies heavily on this Court’s decision in Raffel v. United States, 271 U. S. 494 (1926).^[4] There, a second trial was required when the first jury failed to reach a verdict. In reliance on his privilege against compulsory self-incrimination, the accused declined to testify at his first trial. At the second trial, however, he took the stand in an effort to refute the testimony of a Government witness. Over objection, Raffel admitted that he had remained silent in the face of the same testimony at the earlier proceeding. Under these circumstances the Court concluded that Raffel’s silence at the first trial was inconsistent with his testimony at the second, and that his silence could be used to impeach the credibility of his later representations. The Government argues that silence during police interrogation is similarly probative and should therefore be admissible for impeachment purposes.

We cannot agree. The assumption of inconsistency underlying Raffel is absent here. Rather, we find the circumstances of this case closely parallel to those in Grunewald v. United States, 353 U. S. 391 (1957), and[*176] we conclude that the principles of that decision compel affirmance here.

II

A basic rule of evidence provides that prior inconsistent statements may be used to impeach the credibility of a witness. As a preliminary matter, however, the court must be persuaded that the statements are indeed inconsistent. 3A J. Wigmore, Evidence § 1040 (J. Chadbourn rev. 1970) (hereafter Wigmore). If the Government fails to establish a threshold inconsistency between silence at the police station and later exculpatory testimony at trial, proof of silence lacks any significant probative value and must therefore be excluded.

In most circumstances silence is so ambiguous that it is of little probative force. For example, silence is commonly thought to lack probative value on the question of whether a person has expressed tacit agreement or disagreement with contemporaneous statements of others. See 4 Wigmore § 1071. Silence gains more probative weight where it persists in the face of accusation, since it is assumed in such circumstances that the accused would be more likely than not to dispute an untrue accusation. Failure to contest an assertion, however, is considered evidence of acquiescence only if it would have been natural under the circumstances to object to the assertion in question. 3A iVigmore § 1042. The Raffel Court found that the circumstances of the earlier confrontation naturally called for a reply. Accordingly, the Court held that evidence of the prior silence of the accused was admissible. But the situation of an arrestee is very different, for he is under no duty to speak and, as in this case, has ordinarily been advised by government authorities only moments earlier that he has a right to remain silent, and that anything he does say can and will be used against him in court.

[*177] At the time of arrest and during custodial interrogation, innocent and guilty alike — perhaps particularly the innocent — may find the situation so intimidating that they may choose to stand mute. A variety of reasons may influence that decision. In these often emotional and confusing circumstances, a suspect may not have heard or fully understood the question, or may have felt there was no need to reply. See Traynor, The Devils of Due Process in Criminal Detection, Detention, and Trial, 33 U. Chi. L. Rev. 657, 676 (1966). He may have maintained silence out of fear or unwillingness to incriminate another. Or the arrestee may simply react with silence in response to the hostile and perhaps unfamiliar atmosphere surrounding his detention. In sum, the inherent pressures of in-custody interrogation exceed those of questioning before a grand jury and compound the difficulty of identifying the reason for silence.^[5]

Respondent, for example, had just been given the Miranda warnings and was particularly aware of his right to remain silent and the fact that anything he said could be used against him. Under these circumstances, his failure to offer an explanation during the custodial interrogation can as easily be taken to indicate reliance on the right to remain silent as to support an inference that the explanatory testimony was a later fabrication. There is simply nothing to indicate which interpretation is more probably correct.

Ill

Our analysis of the probative value of silence before police interrogators is similar to that employed in Grunewald v. United States, supra. In that case a witness before a grand jury investigating corruption in the Inter[*178] nal Revenue Service declined to answer a series of questions on the ground that the answers might tend to incriminate him. The witness, Max Halperin, was later indicted for conspiracy to defraud the United States. At trial he took the stand to testify in his own defense, and there responded to the same questions in a manner consistent with innocence. On cross-examination the prosecutor elicited, for purposes of impeachment, testimony concerning the defendant’s .earlier invocation of the Fifth Amendment on the same subject matter. The Court framed the issue of Halperin’s prior silence as an evidentiary problem and concluded that the circumstances surrounding Halperin’s appearance before the grand jury justified his reliance on the Fifth Amendment, imposed no mandate to speak, and presented valid reasons, other than culpability, for deferring comment. The Court ruled that Halperin’s prior silence was not so clearly inconsistent with his later testimony as to justify admission of evidence of such silence as evidence of a prior inconsistent “statement.”

In Grünewald the Court identified three factors relevant to determining whether silence was inconsistent with later exculpatory testimony: (1) repeated assertions of innocence before the grand jury; (2) the secretive nature of the tribunal in which the initial questioning occurred;^[6] and (3) the focus on petitioner as potential defendant at the time of the arrest, making it “natural for him to fear that he was being asked ques[*179] tions for the very purpose of providing evidence against himself.” 353 U. S., at 423.

Applying these factors here, it appears that this case is an even stronger one for exclusion of the evidence than Grünewald. First, the record reveals respondent’s repeated assertions of innocence during the proceedings; there is nothing in the record of respondent’s testimony inconsistent with his claim of innocence. Second, the forum in which the questioning of Hale took place was secretive and in addition lacked such minimal safeguards as the presence of public arbiters and a reporter, which were present in Grünewald. Even more than Halperin, respondent may well have been intimidated by the setting, or at the very least, he may have preferred to make any statements in more hospitable surroundings, in the presence of an attorney, or in open court. Third, Hale’s status as a “potential defendant” was even clearer than Halperin’s since Hale had been the subject of eyewitness identification and had been arrested on suspicion of having committed the offense.

The Government nonetheless contends that respondent’s silence at the time of his arrest is probative of the falsity of his explanation later proffered at trial because the incentive of immediate release and the opportunity for independent corroboration would have prompted an innocent suspect to explain away the incriminating circumstances. On the facts of this case, we cannot agree. Petitioner here had no reason to think that any explanation he might make would hasten his release. On the contrary, he had substantial indication that nothing he said would influence the police decision to retain him in custody. At the time of his arrest petitioner knew that the case against him was built on seemingly strong evidence — on an identification by the complainant, his flight at that time, and his possession of $158. In these cir[*180] cumstances he could not have expected the police to release him merely on the strength of his explanation. Hale’s prior contacts with the police and his participation in a narcotics rehabilitation program further diminished the likelihood of his release, irrespective of what he might say. In light of the many alternative explanations for his pretrial silence, we do not think it sufficiently probative of an inconsistency with his in-court testimony to warrant admission of evidence thereof.

IV

Not only is evidence of silence at the time of arrest generally not very probative of a defendant’s credibility, but it also has a significant potential for prejudice. The danger is that the jury is likely to assign much more weight to the defendant’s previous silence than is warranted. And permitting the defendant to explain the reasons for his silence is unlikely to overcome the strong negative inference that the jury is likely to draw from the fact that the defendant remained silent at the time of his arrest.^[7]

As we have stated before: “When the risk of confusion is so great as to upset the balance of advantage, the evidence goes out.” Shepard v. United States, 290 U. S. 96, 104 (1933). We now conclude that the respondent’s silence during police interrogation lacked significant probative value and that any reference to his silence under such circumstances carried with it an intolerably prejudicial impact.

[*181] Accordingly, we hold that under the circumstances of this case it was prejudicial error for the trial court to permit cross-examination of respondent concerning his silence during police interrogation, and we conclude, in the exercise of our supervisory authority over the lower federal courts, that Hale is entitled to a new trial.

The judgment below is

Affirmed.