KELLY, J.
This is a claim of ineffective .assistance of counsel. Defendant was convicted by a jury on three counts of criminal sexual conduct involving two sisters. His convictions were based largely on testimony of the older girl who stated that defendant had severely injured her during an incident of sexual misconduct. Defendant maintained that he was innocent and that [480]*480the injury this girl sustained was caused by a bicycle accident, as she had originally related.
Defendant’s counsel failed to adequately interview members of the family who were present on the day of the incident. He did not determine if in fact the alleged bicycle accident had caused the older girl’s injury. On the basis of well-established law, we hold that counsel’s failure to investigate and substantiate defendant’s primary defense was not a strategic decision, erroneous only in hindsight. It was a fundamental abdication of his duty to conduct a complete investigation, and it restricted his ability to make reasonable professional judgments and put forth his case. As a consequence, defendant was deprived of a substantial defense and of the effective assistance of counsel. We reverse the convictions and remand the case for a new trial.
i
The facts in this case were developed at trial and through several posttrial hearings before the trial court.1 A detailed understanding of them and when they were presented is necessary to fully evaluate the appeal. At the time of the alleged incident, defendant was living with his girlfriend at her parents’ home. The sisters are his girlfriend’s nieces. They alleged that defendant sexually abused them on two occasions. The first time was at a birthday party for their grandfather, when the older of them was about eight years old. She alleged that defendant forced her to have intercourse with him. The second allegation was that defendant sexually touched both girls in a closet about a year later.
[481]*481On the day of the first alleged incident, the older girl was severely injured. She suffered a tear from the rear of her vaginal opening to her anus. She told her family and her treating doctor that she had injured herself in a bicycle accident. The examining doctor described the injury as a “clean” tear, consistent with a straddle injury, rather than a ragged tear consistent with abuse. This doctor prepared an initial report of his examination that included the older girl’s statements. He prepared a subsequent report that concluded that, alternatively, her injury could have been caused by sexual abuse.
After the second alleged incident, which occurred about a year later, the older girl told a friend that defendant had had intercourse with her. The friend told her mother, who called child protective services. In connection with the resulting investigation, the girls’ father took them to a second doctor. During the older girl’s examination by this doctor, she said that defendant had raped, then threatened her, demanding that she fabricate the bicycle accident to explain her injury. This doctor also prepared a report of her examination of the complainants, which she provided to the police officer who was investigating the alleged abuse.
The prosecutor proceeded to trial on the theory that the bicycle accident was a fabrication. The older girl testified that her injury was the result of sexual abuse by defendant. She testified that there had never been a bicycle accident at all. The prosecutor’s evidence also included testimony by both examining doctors and the investigating officer. In closing argument, the prosecutor emphasized that defendant had presented no eyewitness testimony to support the occurrence of a bicycle accident.
[482]*482Before trial, defense counsel had available to him at least three sources of information about the charges against defendant.2 (1) He had a copy of the first doctor’s first report, and knew about or had a copy of his second report. (2) He knew about and possibly had a copy of the second doctor’s report. (3) He had a list given him by defendant of at least twelve people associated with the girls or defendant to interview for information or as witnesses.
Defense counsel’s investigators interviewed only two or three of these people. None of them had seen the alleged bicycle accident. Counsel did not direct his investigators to inquire whether the people interviewed could name anyone who had seen it or knew more about it. Consequently, he failed to learn that there were eyewitnesses. Two of the sisters’ cousins could have testified that, on the day of the alleged incident, they saw the older girl injure her genital region in a bicycle accident.
Defense counsel proceeded to trial on a three-pronged theory: (1) defendant did not commit the crimes, if they even occurred; (2) the injury to the older girl was the result of the bicycle accident; and (3) this girl habitually made up things. He argued that, despite the absence of eyewitness testimony, several witnesses said they had heard about the accident, not from the older girl, but from her brother. The jury convicted defendant as charged.
Defense counsel learned of the potential eyewitnesses at the time of sentencing. The girls’ aunt approached counsel and told him that her sons, their cousins, had witnessed the accident. Defense counsel’s [483]*483motion to reopen proofs, presumably to present newly discovered evidence, was denied.
Defendant then retained different counsel who sought a new trial on the basis of newly discovered evidence.3 During a lengthy hearing in the trial court, the cousins testified that they witnessed the older girl injure herself in the bicycle accident. However, the trial court determined that the exculpatory evidence would have been merely cumulative.
On direct appeal, the Court of Appeals found that counsel could have discovered and produced the evidence at trial using reasonable diligence. Hence, defendant was not entitled to a new trial on the basis of newly discovered evidence. But, the Court did find that the evidence was material and not cumulative. It remanded the case for a Ginther4 hearing regarding whether counsel had been ineffective for failing to discover or present the evidence. Unpublished opinion per curiam, issued May 16, 2000 (Docket No. 214941).
By the time the Ginther hearing was held before the trial court, the two cousins only vaguely recalled the incident. This is not surprising considering that the alleged accident had occurred more than five years earlier when they were about ten and six years old. The trial court ruled that the evidence was not sufficiently [484]*484probative to support a determination that counsel was ineffective for failing to ascertain and introduce it. It appears that the trial court’s decision was based on the fact that the witnesses were unable to remember the incident clearly at the time of the Ginther hearing.
Defendant again appealed. The Court of Appeals, apparently analyzing only the Ginther hearing testimony, agreed with the trial court that the evidence “would not have been of substantial benefit to the defense.” Unpublished memorandum opinion of the Court of Appeals, issued May 1, 2001 (Docket No. 214941).
Defendant sought leave to appeal in this Court, requesting a new trial on alternate theories: either the eyewitness testimony of the bicycle accident was newly discovered evidence, or defendant had been denied the effective assistance of counsel by counsel’s failure to produce eyewitnesses at trial. At oral argument before this Court, defense counsel conceded that this evidence would have been discoverable with reasonable diligence and, therefore, was not “newly discovered.” We consider only whether defendant was deprived of the effective assistance of counsel.
ii
Whether a person has been denied the effective assistance of counsel is a mixed question of fact and constitutional law. A judge must first find the facts, then must decide whether those facts establish a violation of the defendant’s constitutional right to the effective assistance of counsel. People v Riley, 468 Mich 135, 139; 659 NW2d 611 (2003). We review a trial court’s findings of fact for clear error. People v LeBlanc, 465 [485]*485Mich 575, 579; 640 NW2d 246 (2002).5 Questions of constitutional law are reviewed de novo. Tolksdorf v Griffith, 464 Mich 1, 5; 626 NW2d 163 (2001).
hi
In People v Pickens,6 this Court adopted the ineffective assistance of counsel standard that the United States Supreme Court established in Strickland v Washington, 466 US 668; 104 S Ct 2052; 80 L Ed 2d 674 (1984). Accordingly, to demonstrate ineffective assistance, a defendant must show that his attorney’s performance fell below an objective standard of reasonableness. The defendant must overcome the presumption that the challenged action could have been sound trial strategy. Id. at 689, see also People v Carrick, 220 Mich App 17, 22; 558 NW2d 242 (1996). A reviewing court must not evaluate counsel’s decisions with the benefit of hindsight. Strickland, supra at 689. On the other hand, the court must ensure that counsel’s actions provided the defendant with the modicum of representation that is his constitutional right in a criminal prosecution.
“[Strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional judgments support the limitations on investigation.. .. [C]ounsel has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Strickland, supra at 690-691.
[486]*486The defendant must show also that this performance so prejudiced him that he was deprived of a fair trial. Pickens, supra at 338. To establish prejudice, he must show a reasonable probability that the outcome would have been different but for counsel’s errors. Strickland, supra at 694. A reasonable probability need not rise to the level of making it more likely than not that the outcome would have been different. Id. at 693. “The result of a proceeding can be rendered unreliable, and hence the proceeding itself unfair, even if the errors of counsel cannot be shown by a preponderance of the evidence to have determined the outcome.” Id. at 694.
A
In this case, counsel’s performance was not objectively reasonable. Defendant was facing three counts of sexual misconduct. Two of them were founded wholly on the sisters’ statements implicating defendant. The third and most serious of them was founded on the older girl’s statements and an underlying physical injury. The best refutation of all the charges would have been strong substantive evidence that the older girl’s injury was caused by something or someone other than defendant. Had that charge been defeated, then the other two would have been greatly weakened, given the questionable credibility of the two girls as witnesses. The development of defense counsel’s trial strategy had to consider these facts. His failure to conduct a more thorough investigation to uncover evidence to support an alternate causation theory was objectively unreasonable.
A sound trial strategy is one that is developed in concert with an investigation that is adequately supported by reasonable professional judgments. Counsel must make “an independent examination of the facts, [487]*487circumstances, pleadings and laws involved ... Von Moltke v Gillies, 332 US 708, 721; 68 S Ct 316; 92 L Ed 309 (1948). This includes pursuing “all leads relevant to the merits of the case.” Blackburn v Foltz, 828 F2d 1177, 1183 (CA 6, 1987).
We evaluate defense counsel’s performance from counsel’s perspective at the time of the alleged error and in light of the circumstances. Strickland, supra at 689. Thus, counsel’s words and actions before and at trial are the most accurate evidence of what his strategies and theories were at trial.
At the Ginther hearing before the trial court on defendant’s claim of ineffective assistance of counsel, defense counsel responded to questioning. He said that his theory had been that the older girl was in the habit of telling lies and could not be trusted. His “main thrust was that this girl was a liar” and he “welcomed” her testimony that she had lied about the bicycle accident. She had been, he theorized, either injured in a bicycle accident or by a sexual assault, but, regardless, was falsely accusing defendant.
Yet, counsel did not think it necessary to be prepared to prove the occurrence of the bicycle accident in order to substantiate his theory that it had caused the injury. He felt that additional witnesses would not be vital. He failed to contact most of the persons whose names defendant had provided for his own defense. He failed to inquire whether anyone in the family had seen and could testify about the fact of the alleged bicycle accident and its role in causing the injury. He failed to act on statements from the witnesses that he did interview that the girls’ brother may have seen the accident.
Justice WEAVER disagrees that defense counsel thought the occurrence of the accident was disputed [488]*488because he testified at the Ginther hearing on his own effectiveness that “ ‘The accident was not disputed. The girl never disputed it.’ ” Post at 520. However, his subjective belief was unreasonable.
Counsel had readily available to him information that should have prompted further inquiries. For example, defense counsel admitted at the Ginther hearing that the first doctor’s initial report said that the injury was not caused by sexual abuse.7 When that doctor testified at trial, he stated that the older girl had told him “[t]hat she was riding a bicycle and slipped and had a straddle injury .. . .” The girl told the second doctor that she had not been injured in a bicycle accident. This doctor testified that the older girl said that
she had been told by this person who abused her not to tell anyone because ... he would hurt her and he threatened her .. . and so she told [the first doctor] that this was a bike accident and this was the story she was told to give him by this abuser.
It is reasonable to infer that the doctors’ testimony was based on their patient histories and the reports they had prepared. That evidence shows that the girl had made conflicting statements about the cause of her injury.
[489]*489The officer investigating the alleged abuse had “asked that a letter be dictated” of the results of the examination, which the doctor did. On cross-examination of this witness, defense counsel referred to a notation he had made in his copy of this doctor’s report, indicating that he had seen it before trial. Hence, defense counsel had seen both reports before trial.
This testimony clearly demonstrates8 that the two doctors’ reports showed conflicting causes of the injury and conflicting statements by the older girl. Hence, defense counsel knew or should have known before trial that the cause of the injury was in question.
Chief Justice CORRIGAN, post at 506 n 2, and Justice WEAVER, post at 519, criticize the majority for relying on the second doctor’s report. We find that it was unreasonable for defense counsel to rely on the older girl’s anticipated testimony at trial to refute the allegation that defendant had caused her physical injury. It would have been unreasonable even if the second doctor’s report had not indicated that the girl was changing her story and even if defense counsel had lacked the report. A central element of his defense was that the girl had falsely accused defendant. It was not reasonable for counsel to rely on part of her testimony to establish an important fact while hoping to show her a liar as to the rest.
Also contrary to Chief Justice Corrigan’s assertions, post at 509, defense counsel acknowledged that it was important to establish that the bicycle accident occurred. He came to this realization in the course of the Ginther hearing. Defense counsel was asked:
[490]*490Q. Would it have heen important for the jury to hear testimony, in your opinion, on behalf of Mr. Grant, that they observed vaginal bleeding from this bicycle accident or...
A. Yeah. If...
Q. ... bleeding in that area?
A. Right. But your question was about witnesses to the accident. You’re, you’re not asking about witnesses to the bleeding. So the answer to the witnesses, the accident, no, that was not important.
Later in the hearing, defense counsel was asked:
Q. Counsel? The issue, as you say, was not the accident. The issue was the cause of the bleeding.
A. Correct.
Q. If you had an eyewitness who was able to not only say, “I saw the accident,” which you say is elementary because it’s, it’s irrelevant. But he can say, “I saw the accident,” and, and “I saw the cause of the bleeding” that occurred from the bicycle accident. Would that type of eyewitness have been important to the defense? That the injury was sustained by the accident and not by criminal sexual conduct?
A. A, a, a civilian eyewitness can say that that’s what caused the bleeding? I don’t think any such thing existed.
Q. Well, let me ask you, sir, if you put a witness on the stand and that witness says, “I saw the little girl riding her bicycle.” “And I saw her get into an accident and I saw her bleeding afterwards.” Would that have been relevant to this defense?
A. If such a witness existed, I guess so, yeah[9]
Despite his later characterization of his decision-[491]*491making as “informed,” we cannot conclude that counsel’s failure to investigate the alleged bicycle accident was in pursuit of a trial strategy, erroneous only in hindsight. People v Johnson, 451 Mich 115, 122-123; 545 NW2d 637 (1996).
Because counsel failed to prepare himself, he failed to appreciate his client’s predicament: without direct evidence of the accident that caused the older sister’s injury, his defense was merely a credibility contest between a little girl and an accused rapist. Witnesses who saw the older sister descend a hill on a bicycle, fall, and return with blood-soaked pants could have provided substantive evidence that abuse did not cause her injury.10 The only evidence that the prosecutor presented to prove these three counts was the testimony and statements of the two girls and the fact of the older girl’s physical injury.
Counsel’s lack of forethought is critical considering that, as he himself opined, in cases like this, the defendant practically has to be proven innocent to be [492]*492acquitted. Given these circumstances, a defense founded solely on credibility was sorely vulnerable to defeat.
We also note that this is not an instance in which counsel failed to discover facts after a reasonable inquiry that would have caused an effective attorney to inquire further. As stated,11 at no time did counsel direct his investigators to ask whether anyone had seen the bicycle accident. Cf. Wiggins v Smith, 539 US 510; 123 S Ct 2527; 156 L Ed 2d 471 (2003) (failure to investigate). His failure to conduct an investigation to determine if known witnesses had direct evidence to substantiate his defense was objectively unreasonable. See Frazier v Huffman, 343 F3d 780, 795 (CA 6, 2003). It is even more so where his witnesses testified that they had heard about the accident from the girls’ brother. He should have recognized that his witnesses could not give substantive evidence of the accident based on another’s out-of-court statements. MRE 802.
Moreover, this is not a case of counsel disregarding one possible, alternate theory of defense in favor of a better one, after finding the first “contradictory, confusing, incredible, or simply poor.” Pickens, supra at 325. As stated above, counsel’s theory was that the girl was a liar and had falsely accused defendant. This was a sound defense strategy.12 Had it been fortified by adequate investigation, it would have shown the weakness in the prosecutor’s case, and it could have made a difference in the verdict. See my discussion beginning at pp 493-497.
[493]*493This case differs from one in which there has been a failure to call witnesses whose potential testimony defense counsel already knows. Cf. People v Johnson, 451 Mich 115; 545 NW2d 637 (1996); People v Corbin, 463 Mich 590; 623 NW2d 884 (2001). Here, counsel did not interview half of the people whom defendant identified as potentially having helpful information.13 He did not know what testimony these witnesses would give. He did not know where they had been or what they had seen.
The fact that defense counsel obtained no substantive evidence of the cause of the older sister’s injury shows that his investigation was incomplete. He relied on the girl’s own, already recanted explanation. His decision not to call as witnesses the individuals identified by defendant was not based on objectively “reasonable professional judgments.” Consequently, his trial strategy was unreasonable under these circumstances.
B
The failure to make an adequate investigation is ineffective assistance of counsel if it undermines confidence in the trial’s outcome. Corbin at 590. Counsel’s failure to investigate his primary defense prejudiced defendant. It adversely affected the outcome, depriving defendant of a fair trial. In light of the evidence presented at trial, there is a reasonable probability that the outcome would have been different.
It was critical to defendant’s theory to show that the older sister had been injured in a bicycle accident. At trial, counsel tried belatedly to establish in the jury’s [494]*494mind the idea that the bicycle accident was real rather than a mere story told by a frightened girl. But, the jury heard no direct evidence that the girl’s injury could have been occasioned by a bicycle accident. Most of the evidence defense counsel attempted to elicit to substantiate the occurrence of the accident was inadmissible as hearsay. The evidence that defense counsel did present concerned the girl’s bleeding, serving only to underscore the severity of her injuries. Later, the prosecutor’s closing argument emphasized the defense’s lack of evidence.
On appeal from the trial court’s denial of defendant’s motion for a new trial, the Court of Appeals correctly held that the cousins’ testimony was not cumulative. It was the best evidence available in support of defendant’s theory. Eyewitness descriptions of the accident would have given independent support to defendant’s theory that the injury was caused by a bicycle accident, not by sexual misconduct.
As the Court of Appeals explained, the girls’ cousins’ testimony “could have transformed a defense theory without any substantiation to a theory supported by observation of eyewitnesses.” Unpublished opinion per curiam, issued May 16, 2000 (Docket No. 214941), p 2. Hence, it was more probative than the older girl’s own earlier statements or the statements of the other witnesses presented at trial, which were admissible only for impeachment. As the Court of Appeals recognized, “[t]his testimony was not corroborative; it would have materially changed the quality, as opposed to the quantity, of the evidence supporting defendant’s theory.” Id.
Had the jury heard the cousins’ testimony about the alleged accident, the nature of the defense would have changed from an unsubstantiated argument to the jury. [495]*495It would have become a direct attack on the factual basis of the prosecution’s primary charge grounded in credible testimony.
The testimony of the two eyewitnesses would have demonstrated that the older girl’s physical injury was the result of a bicycle accident, not sexual abuse. It would have greatly undermined the older girl’s credibility and strongly suggested that she was fabricating horrific stories about defendant.
Had the eyewitnesses testified, the prosecutor’s only remaining evidence of these three counts would have been the testimony of the younger girl.14 After hearing the older sister’s other claims, which were fanciful,15 the jury reasonably would have disbelieved the younger sister’s allegations. She might not have testified.16
The trial court considered the hearsay evidence that was presented at trial about the accident and concluded that additional evidence of the same nature would have been merely cumulative. It failed to consider the trial evidence in favor of defendant when it determined whether there is a reasonable probability that the outcome would have been different. Strickland at 694.
For instance, the older girl “was not crying” when she was examined by the first doctor and she was not “afraid.” She was “less nervous than most kids that [496]*496age” and was not “particularly under stress or nervous.” The doctor was later asked:
Q. And the observations that you made were consistent with [the older girl’s] report of an accident from a fall on a bicycle. Isn’t that correct?
A. Correct.
The doctor concluded, on the basis of the girl’s calm emotional state and the physical characteristics of the injury, that sexual abuse was not involved.
The younger girl testified that she and her sister voluntarily “sat on [defendant’s] lap” after one of the alleged CSC-II incidents. They also failed to tell their father or uncle what had allegedly just occurred, although they were there with defendant and the girls. Defendant’s girlfriend testified that he had “never done anything” to the girls. The older girl was “always hanging around with [defendant] and sitting on his lap” and never acted afraid of him. The girls’ grandfather also testified that the attitude of the girls towards defendant never changed.
At the Ginther hearing, the trial court failed to recognize that the question was not whether the cousins’ testimony was probative. The question was not, as Chief Justice CORRIGAN implies in her dissent, post at 501-502, whether the evidence was sufficient to allow a reasonable juror to find guilt “beyond a reasonable doubt.” People v Gonzalez, 468 Mich 636, 640; 614 NW2d 78 (2003). If that standard obtained at a Ginther hearing, an ineffective assistance of counsel claim would fail in almost every instance. The question was whether there was a reasonable probability that the outcome of the trial would have been different had defense counsel adequately investigated the facts before developing his strategy.
[497]*497After the Ginther hearing, when denying defendant’s motion for a new trial, the trial court improperly relied on counsel’s expertise and performance in past cases to evaluate his performance in this case. It noted that defendant’s counsel was experienced in criminal defense work. The dissent succumbs to the same mistake. It is irrelevant that counsel “expended twice his normal resources on this case . . . .” Post at 505. When defense counsel agreed to represent defendant, he committed himself to conducting an adequate investigation of the case. The resources he devoted to other cases are irrelevant to assessing the performance of his duties in this case.
If the eyewitnesses had testified, the older sister’s testimony that she was injured by sexual abuse would have been refuted. This would have seriously impeached her testimony regarding the other incidents of abuse that allegedly occurred more than a year later. It would have corroborated the testimony that defendant had a positive relationship with the girls.
Considering the evidence admitted for and against defendant, there is a reasonable probability that defendant would not have been convicted as charged. The trial court failed to appreciate that counsel’s failure to investigate and substantiate the defendant’s primary defense was a fundamental abdication of counsel’s duty to conduct a complete investigation. It deprived his client of a substantial defense. Consequently, we find that defendant was deprived of the effective assistance of counsel. Because his convictions are not founded on a fair trial, they cannot stand.17
[498]*498IV
In conclusion, defense counsel failed to investigate and substantiate defendant’s primary defense. There is a reasonable probability that the result of this trial would have been different had the evidence in question been presented. This failure was not a strategic decision, erroneous only in hindsight.
We hold that counsel’s failure to conduct a complete investigation was a fundamental abdication of duty that prejudiced defendant, depriving him of a fair trial. Accordingly, the convictions are reversed and the case remanded for a new trial because of the ineffective assistance of defendant’s counsel.
CAVANAGH, J., concurred with KELLY, J.