Petitioner appeals from a judgment denying him post-conviction relief from his convictions for sexual abuse, unlawful sexual penetration, and sodomy. Petitioner argues that the post-conviction trial court erred in concluding that the failure of petitioner’s trial counsel to introduce evidence of the child victim’s recantation before trial in his underlying criminal prosecution did not result in prejudice sufficient to authorize post-conviction relief. The state responds that, because the evidence of petitioner’s guilt was overwhelming, the omission of the recantation evidence did not have a tendency to affect the result of that trial. We are bound by factual findings of the post-conviction trial court if they are supported by evidence in the record. Krummacher v. Gierloff, 290 Or 867, 869, 627 P2d 458 (1981); Martinez v. Baldwin, 157 Or App 280, 282, 972 P2d 367 (1998), rev den 329 Or 10 (1999). We review the post-conviction trial court’s legal conclusions for errors of law. ORS 138.650; ORS 138.220; Ashley v. Hoyt, 139 Or App 385, 391, 912 P2d 393 (1996). We reverse and remand.
We summarize the following facts from the post-conviction record and the post-conviction trial court’s findings. Petitioner and the child’s mother began dating in March 1995, when the child was two and one-half years old. That summer mother began attending a community college. Petitioner often babysat the child, and at times the child spent the night with petitioner. In early December, petitioner, mother, and the child moved into an apartment together. On December 27, the child accompanied her mother to a gynecological examination, and the child remained in the room while mother was examined by a nurse. As the examinátion began, the child approached the nurse, Linda Rodgers, and asked if the nurse was “going to be touching my mother” and “[a]re you going to hurt my mother?” Rogers was concerned that the child’s questions revealed a “prurient interest” and talked with mother about her concerns. Rodgers also referred mother to a counseling center. Mother testified that later that day she asked the child if anyone had ever touched her in a way that she did not like, or in a way that hurt her, and *142the child responded that “[petitioner] does.” Mother continued to cohabit with petitioner until mid-January 1996, when, according to mother, they decided to go their separate ways. However, they continued to live in the same residence for several weeks longer.
In late January and early February 1996, mother took the child to the counseling center recommended by Rodgers, where both mother and the child were interviewed by counselors. Based on statements made by mother and the child, a counselor at the center contacted the Children’s Services Division (now the Department of Human Services). On February 2, Detective Zeliff received a call from a child-welfare worker regarding the alleged abuse. On February 5, Zeliff interviewed the child at the counseling center. The interview was videotaped.1 During the interview, the child, using anatomically correct dolls, demonstrated how petitioner abused her and told the detective about the abuse.
On February 29, mother took the child to Dr. Candelaria for a physical examination to look for evidence of sexual abuse. The child refused to disrobe and would not cooperate, so Candelaria asked one of the female nurses to conduct a genital examination at the child’s next well-baby checkup. In April, nurse Nina Lara conducted a genital examination of the child and concluded that the child’s hymen did not appear to be intact. In a subsequent examination by both Lara and Candelaria, Candelaria observed that the hymen was “septate.” Candelaria concluded that the area was “definitely abnormal or definitely suspicious,” and he opined that it was caused by a penetrating injury. During the examination, the child told Candelaria that a person named “Mark” — which is petitioner’s first name — had touched her genital area. Mother testified that, after the exam by Candelaria and Lara, the child ran around the house chanting “touches my pee-pee.” On another occasion in August 1996, mother testified that she observed the child *143“masturbating” in the bathtub while explaining that petitioner had done that to her while she was in bed.
Petitioner was indicted in September 1996 on two counts of first-degree sexual abuse, ORS 163.427, one count of first-degree unlawful sexual penetration, ORS 163.411, and one count of first-degree sodomy, ORS 163.405, for acts committed during 1995 and 1996. The child was approximately two years old when the abuse allegedly began and almost five years old at the time of petitioner’s trial in 1997. Before trial, the court conducted a hearing to determine whether the child was competent to testify. During the hearing, the following colloquy occurred involving petitioner’s trial counsel and the child:
“Q. And do you remember being touched? Do you remember being touched, honey?
“A. (No audible response.)
“Q. No? Do you? You don’t remember? Then how do you know that he did that? Did somebody tell you that he touched you? You’re shaking your head yes. Who told you that he touched you?
“A. Mom.
“Q. Your mom? And when did your mommy tell you this?
“A. I don’t know.
“Q. You don’t know? And he really didn’t touch you, did he? You’re shaking your head, no, he didn’t? It’s just your mommy told you he did? You’re shaking your head yes. Do you remember when you lived with this Mark [referring to petitioner] over here, the man without the beard? You do? You’re shaking your head yes. Do you remember how long you lived with him? Time’s a tough concept for a little kid, isn’t it? Yeah, you’re shaking your head yes. Did you live with him a long time or a short time?
“A. Long time.
“Q. A long time. Was he nice to you? No? Did he spank you?
“A. He was bad.
“Q. Pardon?
*144“A. He was bad.
“Q. He was bad? Did your mommy tell you he was bad? You’re shaking your head yes? Do you remember him being bad? Do you remember anything he did that was bad, in your own mind? Or do you remember him as being good? Could you try to answer? You don’t know the answer, honey? You can say I don’t know. She shook her head no.”
On redirect examination, the prosecutor asked the child more questions about the abuse, but the child began crying and asked for her mother almost immediately. The court then called a recess and allowed the child to leave the witness stand and return to her mother. After reconvening, the court ruled that the child was not competent to testify because she was easily led and had a “very spotty” memory.
The case was tried to a jury without testimony from the child. Rodgers testified for the state about the child’s “prurient interest” in her mother’s gynecological examination and that she was concerned that the child may have been abused. The state also called as witnesses mother, Zeliff, Candelaria, and Lara. All gave hearsay testimony as to statements made to them, or in their presence, by the child regarding sexual abuse by petitioner. In addition, Zeliff testified regarding the child’s use of anatomically correct dolls during his interview with her, and Candelaria and Lara testified about what they believed were physical injuries to the child’s genital area. Mother also testified that the child’s behavior changed during the period that petitioner babysat for her and while she and the child lived with petitioner.
The defense theory at the underlying trial was that, in retaliation for petitioner’s breaking up with her, mother fabricated the abuse allegations and instructed the child to lie to others about the abuse. Petitioner’s trial counsel presented, among other evidence, expert testimony casting doubt on Candelaria’s medical opinion, as well as expert testimony regarding the unreliability of witness recall, especially child witnesses. Petitioner also testified in his own defense and denied ever touching the child in a sexual manner. Petitioner’s trial counsel did not offer the child’s pretrial hearing statements. Petitioner ultimately was convicted of each of the charged offenses. We affirmed without opinion. *145State v. Horn, 156 Or App 396, 972 P2d 1227 (1998), rev den 328 Or 464 (1999).
In his petition for post-conviction relief, petitioner alleged, among other grounds for relief, that his trial counsel was constitutionally inadequate for failing to offer into evidence the pretrial “recant[ation]” of the child. The post-conviction trial court concluded that, although trial counsel was inadequate in failing to offer the evidence, the omission did not prejudice petitioner:
“Has petitioner [shown], however that ‘there is a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different’? In this case, there appears to have been a large amount of circumstantial and testimonial evidence. It would seem nearly impossible to believe that the recantation presented, especially when attacked by the prosecution, would have had a reasonable probability or tendency to change the result of the proceeding in the face of what was otherwise overwhelming evidence. No prejudice shown.”
On appeal, petitioner assigns error to that conclusion. He argues that, although the post-conviction trial court “used the words of the appropriate standard, ‘would have a tendency to refute the prosecution,’ the context of the court’s analysis makes it clear that the post-conviction court would require more than a mere ‘tendency.’ ” He also argues that the state’s case was neither “overwhelming,” as suggested by the post-conviction trial court, nor was it even “strong.” Petitioner asserts that the state’s medical evidence was, at most, only “lightly probative,” and the videotape along with the testimony of Zeliff was, at best, “equivocal and inconsistent.” According to petitioner, the strongest evidence against him consisted of the hearsay accusations of the child. Petitioner contends that the child’s pretrial recantation would have impeached not only mother’s testimony, but also the child’s accusations introduced through the testimony of other witnesses. Petitioner argues that, because the trial boiled down to a credibility contest, the omission of potentially impeaching evidence necessarily had a “tendency to affect the result of the prosecution.” The state responds that the recantation evidence was weak in that it was elicited through leading *146questions and that the evidence supporting petitioner’s conviction was overwhelming. Therefore, the state insists, the omission did not have a tendency to affect the result, and petitioner was not prejudiced.
In order to obtain post-conviction relief for inadequate assistance of counsel under Article I, section 11, of the Oregon Constitution,2 petitioner must show, “by a preponderance of the evidence, facts demonstrating that trial counsel failed to exercise reasonable professional skill and judgment and that petitioner suffered prejudice as a result.” Trujillo v. Maass, 312 Or 431, 435, 822 P2d 703 (1991); Bogle v. Armenakis, 172 Or App 55, 59, 18 P3d 390 (2001). To prove prejudice, petitioner must establish that his trial counsel’s deficient exercise of skill and judgment had a “tendency to affect the result of the prosecution * * Stevens v. State of Oregon, 322 Or 101, 110, 902 P2d 1137 (1995) (emphasis in original) (quoting Krummacher, 290 Or at 883); Bogle, 172 Or App at 64. The state does not cross-assign error to the post-conviction trial court’s determination that petitioner’s trial counsel did not exercise reasonable professional skill and judgment in failing to introduce evidence of the recantation. Accordingly, we focus solely on petitioner’s contention that he was prejudiced by the omission.
It is clear from the court’s written decision that it concluded that the recantation evidence would not have had a tendency to affect the result of the proceeding. However, petitioner quarrels with the court’s interpretation and application of the “tendency’ standard. Petitioner contends that the court mistakenly posited that, in order for the failure to introduce the evidence to have been prejudicial, there must have been a “reasonable probability” that the error affected the result of his prosecution. It is sufficient, according to petitioner, that but for his trial counsel’s failure to offer the evidence, the outcome of the case could have been different. The state replies that “tendency” means more than simply “any *147possibility,” but it acknowledges that the Supreme Court has held that the standard requires a lower measure of proof than “likelihood.” See Stevens, 322 Or at 110 n 5.
The Supreme Court first used the phrase “tendency to affect the result of the prosecution” in Krummacher. 290 Or at 883. It concluded that “only those acts or omissions by counsel which have a tendency to affect the result of the prosecution can be regarded as of constitutional magnitude.” Id. Although the court did not elaborate on its conclusion, it held that the petitioner’s trial counsel’s failure to object to a leading question did not have a tendency to affect the outcome of the prosecution because the omission was “inconsequential.” Id.
In Stevens, the petitioner established that his trial counsel in his underlying prosecution for rape had failed to investigate adequately and discover witnesses who would call into question the complaining witness’s credibility. This court affirmed the trial court’s denial of post-conviction relief on the ground that the petitioner had not proved “substantial [ ] prejudice.” Stevens v. State of Oregon, 129 Or App 533, 537, 879 P2d 893 (1994). We described “substantial[ ] prejudice” as omissions by a trial counsel that “would have affected the outcome of the case.” Id. (emphasis added). We held that the omissions likely had not affected the outcome of the prosecution. The Supreme Court reversed, concluding that the petitioner’s counsel’s errors “denied [the] petitioner highly valuable impeaching evidence from disinterested witnesses that would have called into question pivotal testimony of the complaining witness.” Stevens, 322 Or at 110. The court phrased the prejudice test in terms of whether the inadequate assistance of counsel “ ‘would have a tendency to affect the result.’ ”3 Id. at 110 n 5 (quoting Krummacher, 290 Or at 833). The court held that the omitted evidence would have a tendency to affect the result, because there were “no other witnesses to the alleged rape and no physical evidence of abuse or trauma,” and the prosecution “necessarily turned on *148the credibility of the complaining witness and of [the] petitioner.” Id. at 108-09.
In Loveless v. Maass, 166 Or App 611, 999 P2d 537 (2000), based on facts similar in many respects to those found here, the petitioner was convicted of sexually abusing a child with whom he had lived for several years. The petitioner’s theory at his criminal trial was that the child’s mother had coached him to say that the petitioner had abused him after the petitioner and the mother ended their marriage. The petitioner’s trial counsel failed to locate a witness who would have testified that the mother had coached the child in various ways. There was no physical evidence of abuse, and the case “turned on whom the jury believed” — the petitioner or the child. Id. at 618. We reversed a judgment denying the petitioner post-conviction relief, reasoning that evidence of the mother’s coaching was “no different from the evidence that the attorney failed to find in Stevens,” id. at 619, and “would definitely have a tendency to affect the outcome.” Id. at 618.
It is readily apparent that the post-conviction trial court’s formulation and application of the prejudice test is not consistent with the Supreme Court’s decision in Stevens. In its written opinion, the court twice cast the test in terms of a probability that, but for counsel’s error, a different result would have followed. That standard requires too much. See Ashley, 139 Or App at 395-96 (rejecting, in light of Stevens, post-conviction trial court’s adoption of “reasonable probability” standard for prejudice). The question, instead, is whether the omitted evidence would have a tendency to affect the outcome of the prosecution.
The state correctly notes that a criminal defense attorneys failure to investigate, discover, or adduce evidence does not always have a tendency to affect the outcome of a prosecution. Instead, the assessment must be made in light of “the totality of circumstances.” Carias v. State of Oregon, 148 Or App 540, 543, 941 P2d 571 (1997).4 Where evidence *149omitted from a criminal trial is not produced in a post-conviction proceeding, or the evidence could not have changed the result of the prosecution, its omission cannot be prejudicial. See New v. Armenakis, 156 Or App 24, 964 P2d 1101 (1998), rev den 328 Or 594 (1999) (holding that, where the substance of an omitted witness’s testimony was not established in the post-conviction proceeding through admissible evidence, and the petitioner’s recital of the witness’s expected testimony demonstrated that it would not, in any event, have corroborated the petitioner’s defense in the underlying prosecution, the omission did not prejudice the petitioner).
We also note that, since Stevens was decided, this court has concluded that a petitioner’s trial counsel’s failure to present evidence was not prejudicial under the Oregon Constitution where there was ‘little likelihood” that the error affected the result of his prosecution. In Gorham v. Thompson, 159 Or App 570, 575-76, 978 P2d 443 (1999), aff'd on other grounds 332 Or 560, 34 P3d 161 (2001), we concluded that omitted expert testimony, even if admissible in a sex abuse prosecution, was of such marginal value that it would not have overcome compelling evidence of the petitioner’s guilt. We stated our conclusion in terms of the existence of ‘little likelihood that [the omitted] evidence would have affected the verdict.” Id. at 576. That formulation has its roots — although we did not explicitly say so in Gorham— in Supreme Court decisions interpreting Article VII (Amended), section 3, of the Oregon Constitution, to require appellate courts to affirm jury verdicts despite the erroneous admission or exclusion of evidence if there is “substantial and convincing evidence of guilt and little likelihood that the error affected the verdict,” that is, the evidentiary error is “harmless.” See, e.g., State v. Walton, 311 Or 223, 230-31, 809 P2d 81 (1991). Thus, the implicit assumption of our holding in Gorham was that an error that meets the constitutional standard for harmless error also would not have a tendency to affect the outcome of a post-conviction petitioner’s prosecution. Stated conversely, an error that is not harmless is prejudicial. The Supreme Court affirmed our decision in *150Gorham, on different grounds, finding it unnecessary to reach the petitioner’s argument that we had applied the wrong prejudice standard in reaching our conclusion. Gorham, 332 Or at 569.5
The circumstances here do not fit within the patterns found in either Carias or New, because the omitted evidence was in fact admitted in the post-conviction proceeding, and it went to the heart of petitioner’s defense theory. Nor, as we now explain, can we say that omission of the evidence was inconsequential or that it would not have had a tendency to affect the result of petitioner’s criminal prosecution.
As petitioner correctly notes, the recantation evidence would have called into question virtually all of the state’s evidence except the medical testimony of Candelaria. He also correctly observes that Candelaria’s testimony was countered by opposing medical testimony that the condition of the child’s hymen was “one of the normal variants” that is not uncommon in young females. Petitioner further asserts that, if the state had attempted to discredit the recantation because it was elicited through leading questions, he could have parried the impact by showing that the child’s statements to Zeliff also were elicited through leading questions. The state responds that petitioner’s argument in that regard is unprovable, because the videotape of Zeliff s interview with the child is not contained in the post-conviction record. Because the other evidence of petitioner’s guilt was overwhelming and petitioner failed to introduce the videotape, the state urges, he did not establish prejudice. We disagree.
In some respects, the state presented stronger evidence of guilt at petitioner’s criminal trial than was adduced in Loveless. Here, for example, there was physical evidence of the alleged abuse. In addition, the testimony of the nurse who conducted mother’s gynecological exam tended— because it was her suspicions, rather than any actions of *151mother’s, that prompted the investigation leading to the child’s disclosure — to discredit petitioner’s theory that the abuse was fabricated.
On the other hand, the state’s medical evidence was countered by a qualified opposing opinion. Furthermore, in keeping with petitioner’s “coaching” theory, the child first disclosed the abuse to mother — not to the nurse — and mother continued to cohabit with petitioner after that alleged disclosure. In short, the corroborating evidence on which the state relied was not so compelling as to foreclose competing interpretations. See Ashley, 139 Or App at 397 (declining, in light of facts susceptible to differing interpretations, to “imply any view as to the ultimate credibility or persuasiveness” of omitted evidence). Although we agree that petitioner’s trial counsel led the child in her pretrial testimony and that much of their colloquy consisted of counsel’s description of the child’s nonverbal responses, the child nevertheless stated that her “mom” had told her that petitioner had touched her. Thus, regardless of whether the videotape showed that Zeliff led the child in eliciting her inculpatory statements, the child’s testimony in response to defense counsel’s questions would have tended to weaken the credibility of her accusations.
In short, petitioner’s defense crucially depended on his casting doubt on the credibility of mother and the child. The child’s recantation testimony therefore was no less important than the evidence that the petitioners’ attorneys failed to locate and produce in Stevens and Loveless. We thus conclude that the evidence was not inconsequential and that its omission had a tendency to affect the result of that prosecution. Accordingly, the trial court erred in concluding that petitioner was not prejudiced.
We appreciate the lengths to which the concurrence has gone to define a minimum threshold for the “tendency to affect the result” test. We make two brief points in response. First, were we considering whether the exclusion of the recantation evidence was harmless, we would be unable to say that there was “little likelihood” that its exclusion affected the result in petitioner’s prosecution. Thus, application of the harmless error standard confirms our conclusion that petitioner was prejudiced. However, we do not hold that *152the “tendency to affect the result” test is equivalent to a harmless error standard. We merely have acknowledged that the Oregon constitutional test for harmless error implicitly was applied in Gorham and that the federal harmless error standard was applied in Davis. The analogy between error that is not “harmless” and error that results in “prejudice” appears to be useful, but it is for the Supreme Court to decide whether further elaboration of the Krummacher / Stevens test is necessary or desirable.
Which brings us to the second point: The concurrence’s analysis demonstrates — but does not resolve — the imprecision inherent in the phrase “tendency to affect the result.” In the end, the concurrence’s three-part formula does not improve on that imprecision, because it does not tell us what “could have” means. 180 Or App at 155 (Byers, S. J., concurring). We know that it requires less than probability but, like “tendency,” it leaves us uncertain about how much is enough. We do not pretend that the “little likelihood” test for harmless error necessarily is more precise; however, it has the advantage of extensive application in a familiar body of case law. No suggested clarification to date — including the concurrence’s proposal — has clearly defined, for the benefit of bench and bar, the minimum threshold required to establish a “tendency to affect the result.” Any further refinement may require case-by-case development.
Reversed and remanded with instructions to enter judgment allowing post-conviction relief.
Although the transcript of the underlying criminal trial was introduced as an exhibit in the post-conviction trial, some of the exhibits from the underlying trial, including the videotape of the interview, were not introduced by petitioner and, therefore, are not part of the post-conviction record.
Article I, section 11, of the Oregon Constitution, provides, in part, that “[i]n all criminal prosecutions, the accused shall have the right * * * to be heard by himself and counsel!.]”
Petitioner does not assign error to the post-conviction trial court’s determination that he failed to demonstrate inadequate assistance of counsel under the Sixth and Fourteenth Amendments to the United States Constitution.
Although the court misquoted Krummacher by inserting the word “would” into its restatement of the prejudice test, there is no indication that the court believed that, in doing so, it had altered the test in any respect.
In Carias, the petitioner argued that his trial counsel’s failure to contact a potentially valuable witness, without more, constituted prejudice. We disagreed, disavowing the suggestion in Mellem v. State of Oregon, 106 Or App 642, 809 P2d *1491348 (1991), that a petitioner is entitled to post-conviction relief on a mere showing of trial counsel’s failure to investigate. Carias, 148 Or App at 546-47.
In Davis v. Armenakis, 151 Or App 66, 72-73, 948 P2d 327 (1997), rev den 327 Or 83 (1998), we applied as a test for prejudice for post-conviction purposes the federal constitutional standard that, to be harmless, error must be harmless beyond a reasonable doubt. See also Walton, 311 Or at 231. Because we resolve this appeal under the Oregon Constitution, we do not consider the parties’ arguments under the federal standard.