Bagada Dionas, the petitioner, was convicted, by a jury, in the Circuit Court for Baltimore City of multiple counts of second degree murder, first degree assault, use of a handgun in the commission of a felony or crime of violence, openly carrying a dangerous weapon, and conspiracy to commit first degree murder. He was sentenced to an aggregate sentence of life imprisonment plus 170 years. The petitioner appealed his convictions to the Court of Special Appeals, arguing, inter alia, that the trial court erred in prohibiting his cross-examination of a State’s witness regarding that witness’ expectation of leniency in a separate pending case should he testify against *101the petitioner.1 The Court of Special Appeals agreed that the trial court erred in that regard. It, however, concluded that the error was harmless, and affirmed the judgment of the Circuit Court. Dionas v. State, 199 Md.App. 483, 23 A.3d 277 (2011). For the reasons provided below, we now reverse the judgment of the intermediate appellate court.
The charges with which the petitioner was indicted, and of which he was convicted, arose from a shooting incident that occurred on the evening of July 15, 2007, in a field on Radecke Avenue in Baltimore, Maryland, during the course of which two men, Wayne White and Maurice White, were shot and killed. Several persons, one of whom was Sean White (“Mr. White”), the brother of Wayne White, one of the victims, witnessed the incident and testified at the defendant’s trial on behalf of the State.
Following the shooting incident, but prior to the petitioner’s trial, Mr. White was sentenced to probation before judgment for charges of attempted distribution of an illegal substance. While on probation, he was charged with, and plead guilty to, possession of a firearm by a minor, receiving a six-month sentence. As a result of his firearm conviction, Mr. White was charged with violation of probation and, therefore, remained incarcerated on that charge even after he had completed his six-month sentence. He sought a continuance of the violation *102of probation (“VOP”) hearing in order to testify at the petitioner’s trial, and he also requested to be released on home detention pending that hearing. The presiding judge in the VOP case granted his request.
Prior to the petitioner’s trial, the State made a motion in limine to bar the petitioner from questioning any of its witnesses regarding prior arrests. Defense counsel objected, arguing that Mr. White had an agreement for leniency with his VOP judge relevant to Mr. WTiite’s potential motive for testifying. The petitioner’s counsel emphasized that, unlike most individuals, Mr. White had been released pending his VOP hearing, and that that same judge was continuing Mr. White’s hearing until after he testified in the petitioner’s trial. The State maintained that there was no plea deal, that Mr. White had requested a release from jail pending a VOP hearing because he was a witness to his brother’s murder and he was scared in jail.- The court agreed with the State that there was no evidence of any quid-pro-quo between Mr. White and the VOP judge, and that, due to his involvement in the petitioner’s trial, Mr. White was released because it would be safer for him to be home.
The petitioner’s subsequent efforts to cross-examine Mr. White with respect to his motivation to testify were rejected. The trial court ultimately ruled that, although Mr. White and the VOP judge had an agreement,2 that agreement was merely that Mr. White testify; she did not tell Mr. White “directly or indirectly that he had to testify one way or another,” “testify truthfully” or “in a manner consistent with the conviction of the defendant.” Therefore, the trial court reasoned that there was no deal that Mr. Wfiiite testify “one way or the other[,] [o]nly that he testify,” and any questioning about the agree*103ment between Mr. White and the VOP judge would be irrelevant.3
Subsequent to Mr. White’s testimony, the petitioner renewed his request to cross-examine Mr. White concerning his motivation to testify, noting that the question was not whether the State had struck a bargain with Mr. White, but whether Mr. White subjectively believed that the would receive a benefit at his VOP hearing as a result of testifying on the State’s behalf at the petitioner’s trial, a belief bolstered by the fact that he was released to home detention prior to his VOP hearing. As to this argument, the court noted that Mr. White had given a statement to the police regarding the petitioner’s case before he appeared in front of the VOP judge.
After a three and a half day trial, the jury began deliberations, which lasted for five and one half days. On the first day of deliberations, the jury requested further instruction on first and second degree murder. On the second day of deliberations, the jury submitted two notes. The first asked: “We have reviewed each indictment individually. We have been unable to reach a unanimous decision on any single count. What should we do at this point?” The second stated: “We the jury have deliberated the verdict to the best of our knowledge and can only agree on two counts.” The court responded by explaining that it is not uncommon for a jury to believe early on in deliberations that a unanimous verdict cannot be reached, and encouraged the jury to “consult with one another and deliberate with a view towards reaching an agreement.” On the fourth day of deliberations, the jury submitted a question regarding whether the use of a handgun in the commission of a crime of violence counts related specifically to each of the enumerated victims, which the trial court went on the record to answer. Finally, on the fifth day of deliberations, the trial court received two more notes from the jury: “What are the consequences when some jurors are not *104cooperating with deliberations, blatantly disregarding the law of the land, and the evidence presented in this case,” and, from Juror # 9, “I did not talk to no one about this case. I am being pick[ed] out. Some one that I do not know ask [sic.] me if we came to a verdict. My verdict did not match their verdict it made the count 11 to 1, on more than one count.” The court responded by reminding the jurors that they swore an oath to deliberate fairly and impartially and to apply the law as the judge explained it to them. The court also gave the jury a long instruction addressing juror safety and determined that the question put to Juror # 9 was “innocuous, and that the communication should not be held against the State or the defendant in any way.” In addition, on the fifth day of deliberations the court asked the jury: “Do you believe that continued deliberations will result in a unanimous verdict?,” to which the jury responded, “So far in our deliberations, some days we have gotten closer to a unanimous verdict and some days we have gotten further away. It is difficult to determine whether more time to deliberate will result in a unanimous verdict.” On the sixth day of deliberations, the jury returned its verdict.
Before the Court of Special Appeals, as we have seen, the petitioner raised the issue of whether the trial court erred by prohibiting the petitioner’s cross-examination of Mr. White regarding his expectation of leniency in his VOP case. The Court of Special Appeals agreed that it did:
“In the present case, the defense similarly sought to question [Mr. White] regarding whether he had any hope that he would receive leniency in his VOP hearing as a result of his testimony in appellant’s case. Given that [Mr. White]’s VOP hearing has been postponed for him to “complete cooperation” in appellant’s case, and that the VOP judge had granted bail pending the hearing and stated that the parties would bring to her attention at the sentencing hearing his participation as a witness in appellant’s case, there was a sufficient factual foundation to support the requested cross-examination. The trial court erred in finding otherwise and in restricting cross-examination of [Mr. *105White] regarding any expectation of leniency in return for his testimony.”
Dionas, 199 Md.App. at 509, 23 A.3d at 292. The intermediate appellate court, however, found the error to be harmless. The court, purporting to apply the harmless error test enunciated in Dorsey v. State, 276 Md. 638, 350 A.2d 665 (1976), with a focus on factors it had previously recognized in Owens v. State, 161 Md.App. 91, 111, 867 A.2d 334, 345-46 (2005) (quoting Delaware v. Van Arsdall, 475 U.S. 673, 684, 106 S.Ct. 1431, 1438, 89 L.Ed.2d 674, 686-87 (1986)), opined:
“Whether such an error is harmless in a particular case depends upon a host of factors, all readily accessible to reviewing courts. These factors include the importance of the witness’ testimony in the prosecution’s case, whether the testimony was cumulative, the presence or absence of evidence corroborating or contradicting the testimony of the witness on material points, the extent of cross-examination otherwise permitted, and, of course, the overall strength of the prosecution’s case.”
Dionas, 199 Md.App. at 510, 23 A.3d at 292-93 (internal citations omitted). Applying these factors, the Court of Special Appeals noted that the State’s case against the petitioner was strong. It reasoned, while Mr. White’s identification of the petitioner was important, it was cumulative; the petitioner was otherwise given ample opportunity to cross-examine Mr. White; and the impact of the petitioner’s proffered cross-examination would have been minimal. Dionas, 199 Md.App. at 513, 23 A.3d at 294. The court explained:
“At oral argument, counsel for appellant argued that, despite the strong evidence against appellant, and the limited impeachment value of the excluded evidence, the error was not harmless. In support, counsel pointed to the length of deliberations and that the jurors had indicated on the second day of deliberations that they could not reach a verdict. Neither appellant nor the State cited any cases addressing whether lengthy deliberations or a note such as that sent by the jury here were relevant to a finding of harmless error.
*106“Addressing first the length of deliberations, courts have disagreed on whether that fact has any bearing in a harmless error analysis, in particular with respect to an analysis of the strength of the State’s case. Compare United States v. Williams, 212 F.3d 1305, 1311 n. 10 (D.C.Cir.) (courts should “hesitate to connect the length of deliberations with the strength of the government’s case”), cert. denied, 531 U.S. 1056, 121 S.Ct. 666, 148 L.Ed.2d 568 (2000), with United States v. Velarde-Gomez, 269 F.3d 1023, 1036 (9th Cir.2001) (en banc) (“Longer jury deliberations ‘weigh against a finding of harmless error’ because lengthy deliberations suggest a difficult case.”); United States v. Varoudakis, 233 F.3d 113, 126 (1st Cir.2000) (same). Even the courts that have looked to the length of deliberations, however, have noted that a lengthy deliberation may result, “not because of indecision, but [from] ‘a diligent and conscientious attempt to evaluate the evidence, and to verify the testimony of different witnesses and to come to a careful and reasoned decision.” Varoudakis, 233 F.3d at 126-27 (quoting Clark v. Moran, 942 F.2d 24, 32-33 (1st Cir.1991)).
“In the present case, the jury deliberated for a period of time approximately five days. Although this is a lengthy period of time, it was in the context of a trial that spanned four days and involved 33 different counts, involving different legal theories. During the deliberation period, the jury sent out multiple notes, asking for definitions of different offenses, clarification regarding the elements of a crime, and information about specific evidence that was introduced. Under these circumstances, the length of deliberations reflects a conscientious attempt to reach a reasonable decision, and it does not weigh against a finding of harmless error.
“Nor do we find that the two jury notes indicating difficulty reaching a unanimous verdict weigh against a finding of harmless error. Both notes were written on the second day of deliberations, the first after less than six hours of deliberations. And, as indicated, the subsequent deliberations, including the notes requesting more guidance regarding the *107nature of the crimes charged, show a diligent attempt to reach a careful decision.
“Given the strong evidence against the appellant, and the limited impact that the cross-examination likely would have had, we hold that the court’s restriction of cross-examination, although error, was harmless. Appellant is not entitled to a new trial on this ground.”
Id. at 512-13, 23 A.3d at 294 (footnotes omitted). Accordingly, the Court of Special Appeals affirmed the judgment of the Circuit Court. Id. at 537, 23 A.3d at 308.
This court granted certiorari to consider the following question:
“Did the lower court err in finding that the improper limitation of cross-examination of a witness’ expectation of leniency was harmless where the jury deliberated for five days after three and a half days of testimony and sent out four notes indicating that they were unable to reach a unanimous verdict or were otherwise having difficult with deliberations.”
Dionas v. State, 422 Md. 352, 30 A.3d 193 (2011).
II.
The parties agree with the Court of Special Appeals that the trial court erred in limiting the petitioner’s cross-examination of Mr. White with regard to his expectation of leniency from the YOP judge. That leaves for resolution only the question of the effect of that error, whether it was, as the intermediate appellate court determined, harmless to the result of the trial. Because the right to cross-examine a witness on matters and facts that are likely to affect his or her credibility is a fundamental concept in our system of jurisprudence, we employ the harmless error analysis when reviewing its violation. Smallwood v. State, 320 Md. 300, 308, 577 A.2d 356, 359 (1990) (citing Van Arsdall, 475 U.S. at 684, 106 S.Ct. at 1438, 89 L.Ed.2d at 686) (“An appellate court must therefore determine ‘whether, assuming that the damaging poten*108tial of the cross-examination were fully realized, ... the error was harmless beyond a reasonable doubt.’ ”).
The harmless error test is well established, and relatively stringent. We stated it in Dorsey v. State:
“[W]hen an appellant, in a criminal case, establishes error, unless a reviewing court, upon its own independent review of the record, is able to declare a belief, beyond a reasonable doubt, that the error in no way influenced the verdict, such error cannot be deemed ‘harmless’ and a reversal is mandated. Such reviewing court must thus be satisfied that there is no reasonable possibility that the evidence complained of — whether erroneously admitted or excluded — may have contributed to the rendition of the guilty verdict.”
276 Md. at 659, 350 A.2d at 678 (footnote omitted). See Smallwood, 320 Md. at 308, 577 A.2d at 360 (quoting Dorsey, 276 Md. at 659, 350 A.2d at 678) (“[A]n error will be considered harmless if the appellate court is ‘satisfied that there is no reasonable possibility that the evidence complained of— whether erroneously admitted or excluded — may have contributed to the rendition of the guilty verdict.’ ”).
“An evidentiary or procedural error in a trial is bound, in some fashion, to affect the delicately balanced, decisional process____It is the impact of the erroneous ruling upon the defendant’s trial and the effect it has upon the decisional process which is of primary concern.” Dorsey, 276 Md. at 657-58, 350 A.2d at 677. Because a criminal conviction must be based upon proof beyond a reasonable doubt in order to satisfy the constitutional requirements of due process, “an appellate court should not arrive at a conclusion about the impact of an error upon a jury verdict[ ] with any less degree of certainty[.]” Id. at 658, 350 A.2d at 677-78. We have similarly stated that the harmless error standard has been and should be carefully circumscribed. Id. at 661, 350 A.2d at 679. What’s more, once error is established, the burden falls upon the State, the beneficiary of the error, to exclude this possibility beyond a reasonable doubt. Hunter v. State, 397 Md. 580, 596, 919 A.2d 63, 72 (2007). See Denicolis v. State, 378 Md. *109646, 658-59, 837 A.2d 944, 952 (2003) (explaining that in order for the State to show that an error was harmless beyond a reasonable doubt, the record must affirmatively show that the error was not prejudicial). For these reasons, we have stated that harmless error review “is the standard of review most favorable to the defendant short of an automatic reversal.” Bellamy v. State, 403 Md. 308, 333, 941 A.2d 1107, 1121 (2008).
We have also explained that the reviewing court must apply the harmless error standard in a manner that does not encroach upon the jury’s judgment. Id. at 332, 941 A.2d at 1121. The court reviewing the trial court error for prejudice has a function distinct from that of the trier of fact. Id. Certain tasks are reserved to the trier of fact, and performance of these tasks must be preserved upon review. In Bellamy, we stated:
“In performing the harmless error analysis, we are not to find facts or weigh evidence. Instead, ‘what evidence to believe, what weight to be given it, and what facts to flow from that evidence are for the jury ... to determine.’ Dykes v. State, 319 Md. 206, 224, 571 A.2d 1251, 1260-61 (1990).... ‘To say that an error did not contribute to the verdict is, rather, to find that error unimportant in relation to everything else the jury considered on the issue in question, as revealed by the record.’ United States v. O’Keefe, 128 F.3d 885, 894 (5th Cir.1997).”
Id., at 332, 941 A.2d at 1121.
In a criminal jury trial, the jury is the trier of fact. For this reason, it is responsible for weighing the evidence and rendering the final verdict. Therefore, any factor that relates to the jury’s perspective of the case necessarily is a significant factor in the harmless error analysis. Thus, harmless error factors must be considered with a focus on the effect of erroneously admitted, or excluded, evidence on the jury. As we have explained, in a harmless error analysis, the issue is not what evidence was available to the jury, but rather what evidence the jury, in fact, used to reach its verdict. See Bruce v. State, 318 Md. 706, 728, 569 A.2d 1254, 1265 (1990); *110Hook v. State, 315 Md. 25, 42, 553 A.2d 233, 242 (1989). We have stated frequently that, where credibility is an issue and, thus, the jury’s assessment of who is telling the truth is critical, an error affecting the jury’s ability to assess a witness’ credibility is not harmless error. See Martin v. State, 364 Md. 692, 703, 775 A.2d 385, 391 (2001) (“[T]he complete denial to petitioner of an opportunity to impeach the witness’s credibility was not harmless error.”); Howard v. State, 324 Md. 505, 517, 597 A.2d 964, 970 (1991) (“In a case that largely turned on whom the jury was going to believe, the improperly admitted evidence of the defendant’s prior conviction may have been the weight which caused the jurors to accept one version rather than the other.”); State v. Cox, 298 Md. 173, 185, 468 A.2d 319, 324-25 (1983) (“Despite some corroborating physical evidence, the prosecution’s case against Cox was based on the testimony of the victim. If she were shown to be unworthy of belief, the jury might well have been unable to conclude that Cox was guilty beyond a reasonable doubt.”).
In criminal jury cases where error has been established, we have considered a number of factors that may have influenced the jury’s perspective as the arbiters of fact. One such factor is the nature, and the effect, of the purported error upon the jury. In State v. Cox, for example, we reviewed a trial judge’s erroneous denial of the defense’s attempt to cross-examine the State’s witness. 298 Md. at 184, 468 A.2d at 324. In explaining the effect of the error from the jury’s perspective, we stated that “[t]he right to cross-examine effectively necessarily includes the right to place the testimony of a witness in its proper setting to fairly enable the jury to judge its credibility.” Id. (citing Smith v. Illinois, 390 U.S. 129, 132, 88 S.Ct. 748, 750, 19 L.Ed.2d 956, 959 (1968); Alford v. United States, 282 U.S. 687, 692, 51 S.Ct. 218, 219, 75 L.Ed. 624, 628 (1931)). We thus held:
Although the jury had ample evidence to convict Cox, we cannot say beyond a reasonable doubt that the court’s error could not have influenced the jury ... If [the witness] were shown to be unworthy of belief, the jury might well have been unable to conclude that Cox was guilty beyond a *111reasonable doubt. The proffered cross-examination, if successful, could have cast sufficient doubt on the prosecuting witness’ credibility to render her unworthy of belief in the mind of at least one juror. Therefore, the error in this case was harmful, and Cox’s conviction must be reversed.
Id. at 185, 468 A.2d at 324-25.
We have also considered the jury’s behavior during deliberations as a relevant factor in the harmless error analysis. See Hunter v. State, 397 Md. 580, 597, 919 A.2d 63, 72-73 (2007). In Hunter, we concluded that jury notes sent over the course of the jury’s deliberations were relevant to the harmless error analysis. Id. In that case, we considered whether, allowing the State to ask the petitioner several “were-they-lying” questions, was harmless error. Id. at 596, 919 A.2d at 72. Among the factors that the petitioner offered to counteract the factors that the State offered, justifying the trial court’s “harmless error” ruling, i.e. that the jury had been instructed that it was the sole judge of credibility and that the State presented “overwhelming” evidence that the petitioner had committed the crime, id. at 596-97, 919 A.2d at 72, were that during its deliberations, the jury submitted four notes to the trial court, three of which asked for additional information or clarification of certain information, and the fourth advised the trial judge that it doubted its ability to reach an unanimous verdict. Id. One of these questions, we concluded, may have been related to a concern the jury had about the truthfulness of the petitioner’s testimony. Id. The other questions, we observed, may have been referring to the conflict between the officer’s and the petitioner’s testimony, and a juror’s concern regarding an element of the crime, possession of stolen property. Id. at 597, 919 A.2d at 73. After considering the factors weighing both for and against a finding of harmless error, we determined that “[w]e are unable to say, beyond a reasonable doubt, that the jury was not affected by the ‘were-they-lying’ questions. Therefore, the trial court’s error in allowing the questions was not harmless.” Id.
*112We take from Hunter that the length of jury deliberations is a relevant factor in the harmless error analysis. This is so because length of jury deliberations provide context, albeit not necessarily conclusive, for the evaluation and understanding of the jury’s findings, and thus, perspective.
This is consistent with authorities from other States and the federal courts. See, e.g. Parker v. Gladden, 385 U.S. 363, 365, 87 S.Ct. 468, 470, 17 L.Ed.2d 420, 423 (1966) (noting that “the jury deliberated for 26 hours, indicating a difference among them as to the guilt of petitioner”); United States v. Sandoval-Gonzalez, 642 F.3d 717, 726 (9th Cir.2011) (“Our doubt that the errors were harmless is heightened by the length of the jury’s deliberations.”); United States v. Howell, 285 F.3d 1263, 1271 (10th Cir.2002) (citing fact that “the jury deliberated for some seventeen to eighteen hours” as weighing in favor of finding that error was not harmless); United States v. Velarde-Gomez, 269 F.3d 1023, 1036 (9th Cir.2001) (quotation omitted) (“longer jury deliberations weigh against a finding of harmless error [because l]engthy deliberations suggest a difficult case.”); Brooks v. United States, 367 A.2d 1297, 1310 (D.C.1976) (“Resolution of [the harmless error] question involves an examination of the nature of the particular evidence, the emphasis placed on it by the court or the parties, the possibility of a compromise verdict, and indicia of the trier’s belief that the case is ‘close’ (such as the length of deliberations and the questions raised by the jury).”). Compare United States v. Williams, 212 F.3d 1305 (D.C.Cir.2000).
Allen v. United States, 837 A.2d 917 (D.C.2003) is also instructive. There, considering whether a trial court’s error in allowing the prosecutor to ask the appellant to comment on the credibility of another witness was harmless, the appellate court concluded:
“Not only, as we have explained, was witness credibility paramount to the jury’s task, but it was an issue the jury found difficult to resolve. It deliberated over several days, sent out two notes saying that it could not reach a decision, and asked for reinstruction several times ... ‘In order to determine whether [an] error had substantial influence we *113must also consider the jury’s actions during deliberations.’ Barron v. United States, 818 A.2d 987, 993 (D.C.2003) (quoting Kotteakos v. United States, 328 U.S. [750], 765 [66 S.Ct. 1239, 1248, 90 L.Ed. 1557, 1567 (1946)]); see also Brooks v. United States, 367 A.2d 1297, 1309 (D.C.1976) (‘Resolution of [the harmless error question] involves an examination of [among other things the] indicia of the trier’s belief that the case is close ... such as the length of deliberations and questions raised by the jury’). The communications from the jury, and its struggle with the case, confirm our own assessment from the record that the issue of whom to believe was not so one-sided as to neutralize the impact of the prosecutor’s questions.”
Id. at 922-23 (footnote omitted).
The Court of Special Appeals relied primarily on the strength of the State’s case, discounting, to a large degree, the jury’s behavior during deliberations. In addition, that court improperly substituted its fact-finding and credibility determinations for those of the jury; it independently, and, in total disregard of the jury’s responsibility as the trier of facts, weighed the evidence produced at trial. As a result, we hold that the intermediate appellate court misapplied the harmless error test.
To be sure, and as we have seen, the Court of Special Appeals did not announce a rule that the length of deliberations and jury behavior during those deliberations are never relevant to the harmless error test, only that, under the facts of this case, those factors do “not weigh against a finding of harmless error.” Dionas, 199 Md.App. at 513, 23 A.3d at 294. For this conclusion, that, in this case, the length of deliberations and jury behavior during deliberations are not a relevant factor in the harmless error analysis, it relied on out-of-state authority, United States v. Williams, 212 F.3d 1305 (D.C.Cir. 2000) and a statement from United States v. Varoudakis, 233 F.3d 113 (1st Cir.2000). In Williams, the D.C. Circuit found the erroneous admission of a police officer’s testimony regarding his experience as a patrol officer to be harmless, in view of the weight of the evidence against the defendant, which it *114considered to negate the error’s impact. 212 F.3d at 1311-12. It then observed: “We should also hesitate to connect the length of deliberations with the strength of the government’s case.” Id. at 1311 n. 10. In Varoudakis, a panel for the First Circuit held that the trial court’s erroneous admission of prior bad acts evidence was not harmless error. 233 F.3d at 127. Nevertheless, after noting that the jury had deliberated for three days, and sent a note to the trial court stating that it was “at an impasse,” the court stated that “[i]n some cases, the jury may deliberate for an extended period not because of indecision, but in a diligent and conscientious attempt to evaluate the evidence, and to verify the testimony of different witnesses and to come to a careful and reasoned decision.” Id. at 126-27 (quotation omitted).
We do not find Williams persuasive and, if anything, Varoudakis supports the petitioner’s position. The harmless error standard applied by the Williams court is different from the standard we apply in Maryland. That court characterized the relevant standard as whether, “with fair assurance, after pondering all that happened without stripping the erroneous action from the whole, that the judgment was not substantially swayed by the error.” Williams, 212 F.3d at 1310 (quoting United States v. Schaffer, 183 F.3d 833, 852 (D.C.Cir.1999), in turn quoting Kotteakos v. United States, 328 U.S. 750, 765, 66 S.Ct. 1239, 1248, 90 L.Ed. 1557, 1566-67 (1946)).4 That is significantly different from the test we apply, whether, the reviewing court can conclude “beyond a reasonable doubt ... that the error in no way influenced the verdict.” Dorsey, 276 Md. at 659, 350 A.2d at 678.
Williams is distinguishable for another reason. Harmless error was not the issue in that case. The statement on which the Court of Special Appeals and State rely was made in response to the dissenting opinion in that case, which expressed concern that, before the trial at issue, an earlier jury had considered the defendant’s case and was unable to reach a *115decision, resulting in a mistrial. The response, in its entirety, was
“[w]e advise caution in assigning critical significance to the failure of a different jury, which heard different evidence and argument, to reach agreement. We should also hesitate to connect the length of deliberations with the strength of the government’s case.”
Williams, 212 F.3d at 1311 n. 10.
The State’s reliance on Varoudakis is even more unavailing. Although it was a harmless error case, viewed in context, it is clear that it does not support the State’s position. This is so because the court in Varoudakis found expressly that the length of jury deliberations and a jury note weighed against a finding of harmlessness. 233 F.3d at 126 (“[T]he three-day length of the jury deliberations, and the jury’s note to the trial court that it was ‘at an impasse’ at the end of the second half-day, weigh against a finding of harmless error.”). Furthermore, the Varoudakis Court based its finding of harmful error on the substance of a note sent by the jury to the trial judge indicating that it might be unable to reach a unanimous verdict. Id. at 127 (“[T]he jury’s ‘impasse’ note reveals uncertainty about [the defendant’s guilt.”) (citing Medina v. Barnes, 71 F.3d 363, 369 (10th Cir.1995)). The juxtaposition of those statements does not mean, and certainly does not establish, that the length of jury deliberations and jury notes are not relevant to the question of harmless error. Indeed, they may be dispositive, although not necessarily in this case.
Moreover, while the conclusion reached by the intermediate appellate court with regard to the jury’s actions in this case is not an unreasonable one and may, in fact, be accurate as a factual matter, it is not the only rational explanation. Hunter and the cases cited by the petitioner make this clear. Indeed, applying those cases produces a completely different result. As important, accepting that the length of deliberations reflects a jury acting diligently in a conscientious attempt to reach a reasonable decision, that does not mean that the jury was not also struggling with the case, that it was not having *116difficulty resolving the case. In other words, one explanation for the length of deliberations and the jury behavior during those deliberations does not negate every other explanation.
We also conclude that the Court of Special Appeals erred by weighing the strength of the State’s case from its own independent perspective, rather than from the perspective of the jury, as our precedents require. See Bellamy, 403 Md. at 332, 941 A.2d at 1121 (citing Dykes v. State, 319 Md. 206, 224, 571 A.2d 1251, 1260-61 (1990)).
As noted, the Court of Special Appeals relied primarily on the strength of the State’s evidence against the petitioner to conclude that the trial court’s error was harmless. Dionas, 199 Md.App. at 511, 23 A.3d at 293. It reasoned, moreover, that Mr. White’s testimony was cumulative, as it was corroborated by two other witnesses who identified the petitioner as the shooter, id. at 512, 23 A.3d at 293-94, and that Mr. White’s testimony was further corroborated by Mr. White’s two prior extra-judicial identifications of the petitioner as the shooter shortly after the shooting, well before Mr. White’s VOP proceedings. Id., 23 A.3d at 293. Accordingly, the intermediate appellate court concluded that, “[gjiven the strong evidence against [the petitioner], and the limited impact that the cross-examination likely would have had, we hold that the court’s restriction of cross-examination, although error, was harmless.” Id. at 513, 23 A.3d at 294.
By analyzing harmless error in this way, the Court of the Special Appeals, in effect, substituted its fact-finding for the jury’s; it was stating that, if it were hearing the evidence, sitting in place of the jury, it would have believed the State’s witnesses and would have convicted the petitioner, regardless of Mr. White’s testimony or the proffered cross-examination relating to his credibility. That conclusion, that the proffered cross-examination likely would have had limited impact, given the strength of the State’s case, was an assumption that could have only been made upon the evidence it would have credited. Were we to adopt this construction of the Dorsey test, harmless error would be determined on an “otherwise sufficient” *117basis: if the evidence is sufficient without the improper evidence, if the jury could have convicted without it, harm could not have resulted.
An “otherwise sufficient” test, however, is a misapplication of the harmless error test. “Simply stating that the court failed to see how the outcome would be different is not the same as the court determining that the error did not influence the verdict.” Perez v. State, 420 Md. 57, 75, 21 A.3d 1048, 1059 (2011). “To say that an error did not contribute to the verdict is, rather, to find that error unimportant in relation to everything else the jury considered on the issue in question, as revealed by the record.” Taylor v. State, 407 Md. 137, 165, 963 A.2d 197, 213-14 (2009) (quoting Bellamy, 403 Md. at 332-33, 941 A.2d at 1121, in turn quoting United States v. O’Keefe, 128 F.3d 885, 894 (5th Cir.1997), in turn quoting Yates v. Evatt, 500 U.S. 391, 403, 111 S.Ct. 1884, 1893, 114 L.Ed.2d 432, 449 (1991), overruled on other grounds, Estelle v. McGuire, 502 U.S. 62, 72 n. 4, 112 S.Ct. 475, 482 n. 4, 116 L.Ed.2d 385, 399 n. 4 (1991)).
In Yates, the United States Supreme Court concluded that the Supreme Court of South Carolina failed to apply the proper standard to determine whether an error was harmless, when it “sought merely to determine whether it was beyond a reasonable doubt that the jury would have found it unnecessary to rely on the unconstitutional presumptions.” Yates, 500 U.S. at 406, 111 S.Ct. at 1894, 114 L.Ed.2d at 450 (quotation omitted). There, the court stated:
“Enquiry about the necessity for reliance, however, does not satisfy all of Chapman’s concerns. It can tell us that the verdict could have been the same without the [error], when there was evidence sufficient to support the verdict independently of the [error]’s effect. But the enquiry will not tell us whether the jury’s verdict did rest on that evidence as well as on the [error], or whether the evidence was of such compelling force as to show beyond a reasonable doubt that the [error] must have made no difference in reaching the verdict obtained. Because the State Supreme Court’s stan*118dard of review apparently did not take these latter two issues into consideration, reversal is required.”
Id., at 407, 111 S.Ct. at 1895, 114 L.Ed.2d at 451.
Accordingly, the proper inquiry upon applying the harmless error test is not a consideration of the State’s evidence apart from Mr. White’s testimony, but whether the trial court’s error was unimportant in relation to everything else the jury considered in reaching its verdict.5 See Bellamy, 403 Md. at 332, 941 A.2d at 1121. In concluding that “the impact of that cross-examination would have been minimal under the circumstances of the case,” Dionas, 199 Md.App. at 511, 23 A.3d at 293, the intermediate appellate court failed to apply the harmless error analysis as articulated by Dorsey.
The trial judge erroneously limited the defense counsel’s cross-examination of a State’s witness regarding an expectation of leniency. In doing so, the trial judge deprived both the petitioner of his right to confront a witness, and the jury of its function to make a discriminating appraisal of Mr. White’s potential bias and the credibility of his testimony. We cannot conclude, beyond a reasonable doubt, that there was no possibility of prejudice resulting from the trial court’s error.
For these reasons also, we cannot adopt the State’s view of this case. The State, cites the Court of Special Appeals’ opinion at length, asking this Court to apply the harmless error analysis without a proper focus on the jury’s perspective. The State, additionally urges this Court against adopting the reasoning of Hunter, 397 Md. 580, 919 A.2d 63, in considering the case before us. Instead, the State argues that Bellamy, *119403 Md. 308, 941 A.2d 1107, demonstrates that jury behavior during deliberations cannot be a factor that weighs in favor of a finding of harmless error. We disagree.
In Bellamy, we considered whether the trial judge’s erroneous exclusion of a statement by a non-testifying witness was harmless. Id. at 335 n. 25, 941 A.2d at 1122 n. 25. In that case, we held that the trial judge erred in excluding a witness’ statement that should have been admitted as an adoptive admission against the State, id. at 331, 941 A.2d at 1120, but found the error to be harmless. Id. at 335, 941 A.2d at 1122. Important to that decision was a jury note sent during deliberations which asked, “Assuming we have found first degree murder ... do we have to find defendant Bellamy pulled the trigger or if he could have aided and abetted and be guilty of [use of a handgun in a crime of violence]?” Id. at 335 n. 25, 941 A.2d at 1122 n. 25. We concluded:
“Although we do not look behind a jury’s general verdict, we note that the jury may have revealed that it found Bellamy guilty based on aiding and abetting in Carter’s murder. During its deliberations, the jury submitted a note to the judge asking for clarification of the instructions, [referencing the contents of the note] Such a question would be unnecessary if they had found that Bellamy actually pulled the trigger.”
Id.
The State relies on the Bellamy Court’s explanation, that its harmless error ruling would have been the same if the jury had not sent the note to the judge, to argue that the jury’s behavior in the present case would have been similarly irrelevant. The Bellamy Court’s explanation, however, did not reflect on the complexity of the case. Although the statement would have directly contradicted the State’s theory that the defendant was the one who shot and killed the victim, it also would have indicated that the defendant initiated the physical attack against the victim, assaulted the victim, and then physically restrained the victim so that a co-defendant could attack. Id. at 334, 941 A.2d at 1122. Therefore, the Court *120reasoned, if the erroneously excluded statement had been admitted, it alone would have been sufficient for a reasonable jury to convict the defendant of aiding and abetting the murder. Id. at 334-35, 941 A.2d at 1122. In this unusual circumstance, the Court stated, “We will find error harmless where the error resulted in the exclusion of further evidence of a defendant’s guilt.” Id. at 335, 941 A.2d at 1122 (emphasis in original).
In Bellamy, we were careful to distinguish that factual setting from the typical harmless error analysis, noting “[i]t is an unusual occasion where the erroneous exclusion of evidence contradicting material elements of the State’s theory of the crime will be found to be harmless error.” Id. at 333, 941 A.2d at 1121. Additionally, we specifically cited United States v. Evans, 438 F.2d 162, 173 (D.C.Cir.1971) (Bazelon, C.J., dissenting), in which Chief Judge Bazelon explained in dissent: “[I]t will be exceedingly rare to find harmless error when the error concerns ... identification testimony[.]” Id. (citation omitted). In a typical harmless error analysis, the court must consider whether the error may have contributed to the guilty verdict. Dorsey, 276 Md. at 659, 350 A.2d at 678. In contrast, because the error in Bellamy excluded “further evidence of the defendant’s guilt,” Bellamy, 403 Md. at 333, 941 A.2d at 1121, the Court found it unnecessary to examine the jury’s note for any indication that the error may have swayed the jury from a not-guilty verdict. See id. at 335 n. 25, 941 A.2d at 1122 n. 25.
The petitioner’s proffered cross-examination, if successful, would not have provided further evidence of petitioner’s guilt. Instead, the cross-examination would have challenged the credibility of the identification testimony supporting the State’s theory of the crime. Therefore, Bellamy does not provide the support that the State seeks.
Furthermore, adopting the State’s argument would result in an unwarranted expansion of the harmless error analysis. We have previously cautioned against expanding the harmless error standard. “[Continued expansion of the harmless error *121rule will merely encourage prosecutors to attempt to get such [improper evidence] in, since they know that, if they have a strong case, such testimony will not be considered to be reversible error [under the harmless error prejudice requirement].” Perez, 420 Md. at 76, 21 A.3d at 1060 (quoting Younie v. State, 272 Md. 233, 248, 322 A.2d 211, 219 (1974)). What is more, expansion of the standard would undermine the public’s trust and confidence in the jury system. Id.
Applying the harmless error standard established in Dorsey, we cannot conclude beyond a reasonable doubt that the trial judge’s error did not influence the jury’s verdict.
JUDGMENT OF THE COURT OF SPECIAL APPEALS REVERSED. CASE REMANDED TO THE COURT OF SPECIAL APPEALS WITH DIRECTIONS TO REVERSE THE JUDGMENTS ENTERED BY THE CIRCUIT COURT FOR BALTIMORE CITY AND TO REMAND THE CASE TO THAT COURT FOR A NEW TRIAL. COSTS IN THIS COURT AND IN THE COURT OF SPECIAL APPEALS TO BE PAID BY THE MAYOR AND CITY COUNCIL OF BALTIMORE.
. This was only one of the arguments that the petitioner presented on appeal. The other arguments, namely that the trial court coerced a jury verdict by issuing improper Allen-type instructions even after the jury had revealed its numerical division, Dionas v. State, 199 Md.App. 483, 514, 23 A.3d 277, 295 (2011), that the trial court erred by failing to voir dire a juror who had indicated that he had been approached by someone about the jury's verdict and felt pressured, id. at 524, 23 A.3d at 300-01, that his separate sentences for murder and conspiracy to commit murder were illegal and should have been merged as a matter of fundamental fairness, id. at 529-29, 23 A.3d at 303, that the evidence before the trial court was not sufficient to sustain his conviction of conspiracy, id. at 531, 23 A.3d at 305, and that the trial court improperly instructed the jury regarding two theories, transferred intent and kill zone, that were inapplicable to the charges at hand, id. at 533, 23 A.3d at 306, were found to be without merit or not preserved for appellate review. These issues have not been raised on this appeal.
. The petitioner acknowledged that the State had not offered Mr. White a deal, but argued that the VOP judge had because a postponement of Mr. White’s VOP hearing for two years was granted after Mr. White's involvement in the petitioner’s case was brought to the VOP judge’s attention. In other words, "[t]he case kept getting postponed because they wanted to ... hear what happened to the murder trial,” and Mr. White received "home detention at the prompting of [the VOP judge].”
. On the same day that cross-examination of Mr. White concluded, Mr. White's VOP hearing was held. At the hearing, the VOP judge inquired whether Mr. White had testified “consistently.”
. This is the same test that was applied in United States v. Varoudakis, 233 F.3d 113, 126 (1st Cir.2000).
. Even if Mr. White’s testimony was entirely extraneous, its introduction, and its questionable credibility, may still raise the risk of influencing the verdict, as this Court similarly explained in Younie v. State:
"The State's contention that it desired to have the tainted evidence come to the jury’s attention merely to place the concededly proper portion of the confession in some form of comprehensible context demonstrates that it was not harmless error to admit it. This is so, as the good evidence standing alone must be sufficient to convict, and we must be convinced beyond a reasonable doubt that the jury was in no way influenced by the bad."
272 Md. 233, 248, 322 A.2d 211, 219 (1974).