REINHARDT, Circuit Judge:
Sharon Ann Rahm appeals her conviction on charges of possession of counterfeit currency and of attempting to pass counterfeit currency, both in violation of 18 U.S.C. § 472. We hold that the district court erred in excluding expert testimony proffered by Rahm, and reverse and remand.
I
On August 3, 1991, Rahm attempted to purchase a three-dollar box of tea from the Canton Bazaar in San Francisco using a counterfeit $100 bill. The manager of the store recognized the bill as counterfeit, and called the authorities. Although the manager told Rahm more than once that there were some problems with the $100 bill, she remained at the store for fifteen minutes or more, waiting for change or the return of the bill. Rahm had twenty-five dollars in genuine cash in her possession, but she never offered to take back the $100 bill and pay for the tea with this other money. The manager testified at trial that Rahm told him she had received the $100 bill from a bank a few days before.
Responding to the manager’s calls, two police officers arrived and arrested Rahm. On searching her purse, the officers recovered ninety-eight additional counterfeit $100 bills. After waiting several hours at the police station, Rahm was questioned by a Secret Service agent. She signed a statement that she had found the ninety-nine bills at a bus stop and that she thought the money was real. Although this written statement includes her asserted belief that the currency was genuine, the agent states that when questioned orally, she refused to answer whether she knew the money was counterfeit.1
After a two-day trial, Rahm was convicted of possession of counterfeit currency and attempting to pass counterfeit currency. Al[1409]*1409though she unsuccessfully sought to call an expert witness to testify in her defense, ultimately Rahm rested following the prosecution’s case, without introducing any evidence in her own behalf. In argument, her counsel questioned the adequacy of the government’s proof that Rahm knew the bills were counterfeit. Knowledge that the currency is counterfeit is an element of both possession and attempting to pass. Lack of knowledge was Rahm’s sole theory of defense.2
During the government’s case, a Secret Service special agent testified to the quality of the counterfeit bills. He explained what distinguished the ninety-nine bills from genuine currency, identifying the presence of microdots in the printing, the absence of colored security fibers in the paper, and the identical serial numbers on multiple bills. The agent further testified that he had seen both higher-quality and lower-quality counterfeit bills successfully passed off as true currency. On cross-examination, the agent agreed that unsuspecting people have accepted, as genuine, phony bills of the same quality as those in Rahm’s purse.
II
Prior to trial, Rahm gave notice, pursuant to Fed.R.Crim.P. 12.2(b), of her intention to introduce the testimony of Dr. Arvalea Nelson, a psychologist, as to her evaluation of Rahm. Rahm intended to offer the testimony in support of her defense that she did not know the bills were counterfeit.
In examining the defendant, Nelson administered two standardized tests, the Wech-sler Adult Intelligence Scale — Revised (WAIS-R) and the Minnesota Multiphasic Personality Inventory (MMPI). The results of the WAIS-R revealed that Rahm’s intelligence was roughly average; however, her scores on two subtests, Picture Completion and Picture Arrangement, were significantly below average. On both subtests, Rahm showed a “consistent tendency ... to overlook important visual details.” Nelson concluded that Rahm’s performance on the two subtests suggested “difficulty with visual perception.” From the MMPI results, Nelson further concluded that Rahm’s overall personality style might be marked by lack of insight.
The government moved in limine to exclude Nelson’s testimony, disputing its relevance and arguing that Nelson’s evaluation revealed no medically-recognized mental disr order or defect. The district court granted the motion in an oral ruling the day before trial. On appeal, Rahm challenges this decision to exclude Nelson’s expert testimony regarding Rahm’s perceptual difficulties.
Ill
Under Fed.R.Evid. 702, an expert may testify “[i]f scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue.” In United States v. Amaral, 488 F.2d 1148, 1153 (9th Cir.1973), we outlined four criteria to determine the helpfulness of expert testimony: 1) qualified expert; 2) proper subject; 3) conformity to a generally accepted explanatory theory; and 4) probative value compared to prejudicial effect. Although we decided Amaral prior to the enactment of the Federal Rules of Evidence, we have continued to apply these four factors in determining the admissibility of expert testimony under Fed.R.Evid. 702. See, e.g., United States v. Miller, 874 F.2d 1255, 1266 (9th Cir.1989); United States v. Christophe, 833 F.2d 1296, 1299 (9th Cir.1987). Cf. United States v. Fleishman, 684 F.2d 1329, 1337 n. 5 (9th Cir.) (reserving issue whether Amaral survives Rule 702 enactment, but noting Sixth Circuit post-enactment adoption of Amaral factors), cert. denied, 459 U.S. 1044, 103 S.Ct. 464, 74 L.Ed.2d 614 (1982).
We have sometimes described our standard of review in the area of expert testimony as “abuse of discretion” and sometimes as “manifest error.” Compare, e.g., United States v. Barker, 942 F.2d 585, 589 (9th Cir.1991) (abuse of discretion) with United States v. Sinigaglio, 942 F.2d 581, 584 (9th Cir.1991) (manifest error). On occasion, we have used both characterizations. See, e.g., United States v. Arvin, 900 F.2d 1385, 1389 (9th Cir.1990) (admission or exclu[1410]*1410sion of expert testimony “is reversible only for abuse of discretion or manifest error”), cert. denied, 498 U.S. 1024, 111 S.Ct. 672, 112 L.Ed.2d 664 (1991); Miller, 874 F.2d at 1266 (“[w]e review ... decisions to admit or exclude expert testimony for abuse of discretion and reverse only for manifest error”); Christophe, 833 F.2d at 1299 (“we reverse only if the district court abused its wide discretion or committed manifest error”).
We note that general adoption of the “abuse of discretion” characterization would bring this area into line with the rest of our law of evidence. See, e.g., United States v. Rohrer, 708 F.2d 429, 432-34 (9th Cir.1983) (employing general standard to address several evidentiary issues, including exclusion of proffered expert testimony). In a recent en banc opinion, we employed the “abuse of discretion” characterization without directly describing our standard of review. See United States v. Aguon, 851 F.2d 1158, 1171 (9th Cir.1988) (en banc) (“[t]here was no abuse of discretion” in excluding proffered expert testimony). The “manifest error” characterization apparently emanates from a Supreme Court decision preceding the judiciary’s efforts to settle on a limited number of review characterizations. See Salem v. United States Lines Co., 370 U.S. 31, 35, 82 S.Ct. 1119, 1122, 8 L.Ed.2d 313 (1962). There appears to be no practical difference between the two verbal formulae, so their vestigial coexistence serves no obvious purpose. Accordingly, it would be sensible to settle upon a uniform practice of characterizing our standard of review as “abuse of discretion” and abiding by it in all future cases. In view of Aguon, we believe it appropriate to adopt that approach, although for purposes of this opinion, it makes no difference which of the two terms — “abuse of discretion” or “manifest error” — we use.
Generally, a district court abuses its discretion when it bases its decision on an erroneous view of the law or a clearly erroneous assessment of the facts. See Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 405, 110 S.Ct. 2447, 2460-61, 110 L.Ed.2d 359 (1990). In the specific area of expert testimony, we have held that a district court errs when it fails to exercise its discretion as the result of applying an erroneous legal principle. See United States v. Awkard, 597 F.2d 667, 670 (9th Cir.) (reversing where trial court erroneously concluded admission of expert testimony was required under appellate precedent), certs. denied, 444 U.S. 885, 100 S.Ct. 179, 62 L.Ed.2d 116, 444 U.S. 969, 100 S.Ct. 460, 62 L.Ed.2d 383 (1979); United States v. Erskine, 588 F.2d 721, 722 (9th Cir.1978) (reversing where exclusion based on erroneous conclusion that capacity to form intent irrelevant). Here, we must examine certain legal principles upon which the district court’s determination appears to rest.
A
In announcing its ruling on the government’s motion to exclude Nelson’s testimony, the district court stated that it did not think that the proffered testimony met the “threshold” for admission under Rule 702. Beyond this conclusory statement, however, the court provided only the sparest of reasons for its ruling. The court did not address specifically any of the Amaral factors or indicate which it felt were not satisfied by the proffered testimony regarding Rahm’s difficulty with visual perception. We therefore cannot be entirely certain of the rationale for the district court’s decision.
Nonetheless, the court appears to have based the exclusion of Nelson’s testimony upon two erroneous legal premises: 1) that admission of psychological testimony requires that the defendant suffer from a mental disorder; and 2) that an expert must be able to state conclusively that the defendant lacked the requisite mens rea in order to present testimony relevant to that issue. Admission under Fed.R.Evid. 702 requires neither mental disorder nor conclusive opinion.
In granting the government’s motion, the district court commented that it had never seen a case where anyone “was entitled to give opinions about past mental state on people who don’t even have mental disorders.” To the extent this comment suggests that expert psychological testimony may only be introduced about those who suffer from a mental “disorder,” it is an incorrect statement of the law. Nothing in Rule 702 or our [1411]*1411precedent implies such a prerequisite to admissibility. The district court erred in seeking to enforce a mental “disorder” requirement.
Importing such a prerequisite into Fed.R.Evid. 702 would flatly contradict Fed. R.Crim.P. 12.2. The latter rule, with which Rahm complied here, requires advance notice whenever “a defendant intends to introduce expert testimony relating to a mental disease or defect or any other mental condition of the defendant bearing upon the issue of guilt.” Fed.R.Crim.P. 12.2(b) (emphasis added). If admission of psychological testimony under Fed.R.Evid. 702 required mental disorder, the reference in Rule 12.2(b) to “other mental condition” would be entirely superfluous. It is not. The advisory committee had testimony relating to “mental conditions” other than “disorders” specifically in mind when it introduced the broader term to Rule 12.2. See Fed.R.Crim.P. 12.2(b) 1983 amendment advisory committee’s note. We must interpret Fed.R.Evid. 702 in concert with our other federal rules; under these rules, the proper ultimate inquiries are whether evidence of a defendant’s mental condition has a “bearing upon the issue of guilt,” Fed. R.Crim.P. 12.2, and whether expert testimony as to that condition “will assist the trier of fact to understand the evidence or to determine a fact in issue,” Fed.R.Evid. 702. Neither inquiry requires determination of whether the condition merits the label “disorder.”
Furthermore, we note that a rule establishing mental “disorder” as a prerequisite to the admission of expert psychological testimony would undoubtedly prove unworkable in an age when the psychological and psychiatric community is far from unanimous in its constantly-evolving conception of what constitutes “disorder.” Even if a settled definition existed, there is no reason to believe that the scientific classification would bear particular relevance to the main admissibility issue under Rule 702 — helpfulness to the jury. A district court may not exclude proffered expert psychological testimony simply because the defendant’s condition does not fit within the expert’s — or the district court’s own — concept of mental “disorder.”
In excluding Nelson’s testimony, the district court apparently was also troubled by the lack of certainty in Nelson’s conclusion regarding Rahm’s inability to recognize that the currency was not genuine. In Nelson’s report to Rahm’s attorney, she stated that Rahm’s perceptual difficulties and lack of insight “could” have resulted in a failure to examine closely any currency she found or to perceive signs of inauthenticity. Focusing upon the conditional “could,” the district court commented that “that’s not testimony that she lacked capacity, that’s testimony that she had the capacity.” By excluding expert testimony on this basis, the district court imposed a nearly impossible standard, supported neither in logic nor in law.
As an initial matter, we note that the admission of expert testimony, the issue involved here, and the admission of a specific expert opinion are separate determinations. Under the federal rules, an expert “may testify ... in the form of an opinion or otherwise." Fed.R.Evid. 702 (emphasis added). Thus, not every expert need express, nor even hold, an opinion with regard to the issues involved in a trial. Indeed, in certain cases, we will not allow an expert to express an opinion as to specific issues even if he or she has formed one.3 Thus, the decision whether to admit expert testimony does not rest upon the existence or strength of an expert’s opinion. Rather, the key concern is whether expert testimony will assist the trier of fact in drawing its own conclusion as to a “fact in issue.”
[1412]*1412The trial court must consider the probativeness of expert testimony in determining its admissibility. Here, the district court focused exclusively upon the proffered expert’s opinion, rather than considering the psychological test results underlying the opinion. In so doing, the court erred. It is the test results and the perceptual difficulties they demonstrate that would have comprised Nelson’s testimony, and the court should have evaluated them in determining admissibility. The strength of an expert’s opinion, to the extent he has one, may be one factor in determining potential helpfulness to the jury, but it is not dispositive. See United States v. Brewer, 783 F.2d 841, 842 (9th Cir.) (definiteness of opinion legitimately considered as one non-dispositive factor in determining admissibility of expert anthropological testimony), cert. denied, 479 U.S. 831, 107 S.Ct. 118, 93 L.Ed.2d 64 (1986). It is not, ordinarily, a suitable proxy for helpfulness of testimony to a jury, and a district court may not fail to consider the underlying testimony and focus exclusively on whatever opinion the expert may offer.
Furthermore, if the district court considers the strength of an expert’s opinion in deciding whether to admit expert testimony, the court may not seek- — or require— conclusiveness. “Absolute certainty of result is not required for admissibility.” Fleishman, 684 F.2d at 1337; see also id. at 1336— 37 (certainty of expert’s opinion goes to weight of testimony, not admissibility).
Certainty is an unreasonable expectation in the realm of expert opinion. Nelson’s use of the conditional “could” in expressing her conclusion is neither unusual nor disqualifying as to her testimony. In any area of science or social science, but particularly in matters of the mind, expecting an expert to reach a conclusion without the slightest doubt as to its accuracy is exceedingly unrealistic. Experts ordinarily deal in probabilities, in “coulds” and “mights.” See, e.g., Erskine, 588 F.2d at 722 (excerpting proffer of expert testimony that defendant “in the realm of probability” committed crime without knowing what he was doing). It is the rare expert who is willing to opine conclusively about a past occurrence. The expert who expresses his opinion with absolute certainty invites skepticism about his candor or his qualifications. With good reason, we wonder whether such an expert is hiding — or missing — some significant factor. Here, Nelson’s conclusion was simply couched in the probability terminology employed by experts.
The district court erred in two respects by excluding Nelson’s testimony because her opinion was not entirely conclusive. First, the court improperly focused upon Nelson’s opinion rather than the underlying proffered testimony, and, second, the court evaluated Nelson’s opinion under an unreasonable conclusiveness standard.
B
Having concluded that the district court based its exclusion of expert testimony on two erroneous conclusions of law, we must determine whether the proffered testimony otherwise satisfies the requirements for admission under Fed.R.Evid. 702. We undertake this independent review because, as noted above, the basis of the district court’s exclusion was not entirely clear, and because, despite the district court’s legal errors, we will not reverse unless Nelson’s testimony was otherwise admissible. See Fed. R.Crim.P. 52(a) (harmless error not cognizable on appeal); United States v. Gonzales, 749 F.2d 1329, 1336 (9th Cir.1984) (independently reviewing for adequate reasons to exclude where basis of trial judge’s exclusion unclear). Cf. Hingson v. Pacific Southwest Airlines, 743 F.2d 1408, 1413 (9th Cir.1984) (no reversible error in civil case unless proffered expert testimony met Rule 702 requirements). We therefore consider each of the Amaral factors for admission of expert testimony.
Under Amaral, we first require that the testimony be offered by a qualified expert. The government did not challenge Nelson’s qualifications to offer expert testimony; nor did the district court intimate any concerns in this regard. As a consequence, there is little in the record as to the psychologist’s qualifications. Nonetheless, we conclude that Nelson’s testimony would satisfy the first Amaral factor. She is apparently a clinical psychologist with a doctoral degree; her intended testimony related to psychologi[1413]*1413cal tests and evaluation. The government would have been free to evaluate critically her further qualifications upon cross-examination or in argument. Any deficits in her qualifications beyond her professional training go to the weight of her testimony rather than to its admissibility. See United States v. Bilson, 648 F.2d 1238, 1239 (9th Cir.1981) (that psychiatrist was not licensed psychologist goes to weight not admissibility of testimony evaluating psychological tests). Nelson’s testimony was not excludable under the first Amaral factor.
Our “proper subject” inquiry has generally focused upon whether the expert testimony improperly addresses matters within, the understanding of the average juror. See Christophe, 833 F.2d at 1299; Amaral, 488 F.2d at 1152-53. We have stated that expert testimony should not “invaden the province of the jury.” United States v. Binder, 769 F.2d 595, 602 (9th Cir.1985). In recent years, noting that helpfulness to the jury is the central concern of Fed.R.Evid. 702, some scholars and courts have rejected this rationale for the exclusion of experts, concluding that jurors might find expert testimony helpful even in relation to subject matters from everyday experience. See United States v. Langford, 802 F.2d 1176, 1183 (9th Cir.1986) (Ferguson, J., dissenting) (discussing rejection of “jury function” rationale in relation to testimony on unreliability of eyewitness identifications); 3 Jack B. Weinstein & Margaret A. Berger, Weinstein’s Evidence § 702[02] (1992) (arguing that “common experience” rationale for exclusion is inconsistent with helpfulness inquiry). Nonetheless, this circuit continues to guard — perhaps too jealously — from expert elucidation, areas believed to be within the jurors’ common understanding.
The government argues that, consistent with this principle, Nelson’s testimony was inadmissible because it would simply have consisted of her opinion about the credibility of Rahm’s “story.” This contention fundamentally misinterprets our precedent. On occasion in the past, we have stated that expert testimony to bolster or impugn the credibility of a witness is properly excluded. See, e.g., Binder, 769 F.2d at 602 (expert testimony on specific child witnesses’ ability to distinguish truth from falsity and reality from fantasy); Rohrer, 708 F.2d at 434 (possible effects of drug usage by testifying witness); Awkard, 597 F.2d at 669-71 (effect of hypnosis on witness ability to recall). But see United States v. Wesson, 779 F.2d 1443, 1444 (9th Cir.1986) (upholding admission of expert testimony that “may have aided the jury in determining the credibility” of victim’s claim). However, Nelson’s proffered testimony was of an entirely different variety.
All defense evidence necessarily supports the defendant’s “story”; that is its very purpose. If a defendant takes the stand to present his story, we could even loosely characterize the testimony of each additional defense witness as bolstering the “credibility” of the defendant’s testimony. Nonetheless, we do not apply rules regarding credibility evidence to the testimony of each of these other witnesses. Here, Nelson’s testimony related to the issue of knowledge, an element of the crimes with which Rahm was charged, and a “fact in issue” under Fed.R.Evid. 702. The testimony certainly would have enhanced Rahm’s “credibility” had she taken the stand, but all evidence that supports a defendant’s actual — or possible — testimony is not credibility evidence. The purpose of Nelson’s testimony was not to comment upon Rahm’s truthfulness in general or with respect to any account Rahm might offer'as to how she came to possess the counterfeit currency. Rather, Nelson’s testimony was intended to establish Rahm’s diagnosed perceptual difficulties, which were relevant to a fact in issue — whether she knew the money was counterfeit. That Nelson’s proposed testimony supports the defense theory is the reason it was proffered; it is not a reason for its exclusion.
Neither is Nelson’s testimony improper because Rahm herself did not testify that she found the currency and did not know that it was counterfeit. The government argues that without other evidence of lack of knowledge, there was no “foundation” for introducing Nelson’s testimony. In effect, the government suggests that Rahm was required to take the stand to present her “story” before she could present defense evidence to “corroborate” it.
[1414]*1414The rule the government suggests would eviscerate several constitutional rights. Rahm, like all criminal defendants, had the right not to testify. By choosing not to testify, she did not forfeit her right to present a defense or to introduce testimony. Rahm had the right to the presumption of innocence. The government was required to prove each element of the crimes charged beyond a reasonable doubt. Defense testimony calling into question the proof of any such element is proper. Here, the government’s proof of knowledge was entirely circumstantial; the prosecution asked the jury to infer knowledge from Rahm’s possession of the ninety-nine bills and her attempt to use one of them for a purchase. Nonetheless, the government would deny Rahm the right to present circumstantial evidence of lack of knowledge and to present her own inferential argument — that perceptual difficulties, coupled with her remaining in the store for fifteen minutes, should lead the jury to infer that she did not know the money was counterfeit. The government is not entitled to such an unprecedented and unfair advantage. Nelson’s testimony was not improper simply because Rahm herself had not testified to her lack of knowledge.
By other measures, Nelson’s testimony also satisfies the second Amaral factor. Nelson would have interpreted the results of psychological tests; we have approved the introduction of expert testimony for such a purpose in the past. See Bilson, 648 F.2d at 1239. The average juror would not be capable of interpreting the WAIS-R or MMPI results. Moreover, Nelson’s proffered testimony related to Rahm’s specific perceptual difficulties, not to general human deficiencies that the jury could understand from their own experiences. Cf. Miller, 874 F.2d at 1267 (affirming exclusion of testimony consisting of generalities as to people’s usual reaction to failing a polygraph); Christophe, 833 F.2d at 1300 (cross-examination and “appeals to the experience and common sense of jurors” sufficient to alert jury to conditions rendering eyewitness identification unreliable). But cf. Binder, 769 F.2d at 602 (distinguishing general testimony on problems common to all children from specific testimony regarding credibility of specific child witnesses). Nelson’s proffered testimony related to a proper subject.
Neither the government nor the district court expressed any reservations as to the testimony’s conformity “to a generally accepted explanatory theory,” the third Amaral factor. Nothing in the record suggests that Nelson’s testimony would not satisfy this third requisite. Nelson’s letter report suggests that the WAIS-R and MMPI are standardized tests in general use in the psychological profession. If so, the tests are generally-accepted procedures, admissible absent discrediting evidence. See United States v. Hendershot, 614 F.2d 648, 654 (9th Cir.1980) (upholding admission of shoeprint against challenge of technique absent presentation of discrediting evidence). Whether Nelson’s conclusions from the test results are reliable is a separate issue; the government could refute Nelson’s opinion through cross-examination and argument, or with its own expert testimony. See United States v. Bowers, 534 F.2d 186, 193-94 (9th Cir.) (whether use of generally accepted procedure resulted in reliable opinion in particular case is question to be explored in cross-examination of expert witness), cert. denied, 429 U.S. 942, 97 S.Ct. 360, 50 L.Ed.2d 311 (1976). The district court’s exclusion of Nelson’s testimony is not sustainable on the basis of the third Amaral factor.
Finally, we believe that the probative value of the proffered testimony was high, while the danger of unfair prejudice was not significant. Nelson’s testimony was relevant to Rahm’s knowledge that the bills were counterfeit. Such knowledge is an element of a violation of 18 U.S.C. § 472. United States v. Palacios, 835 F.2d 230, 232 (9th Cir.1987). The Secret Service special agent’s testimony at trial indicated that the quality of the counterfeit bills possessed by Rahm was roughly average. Moreover, the agent specifically agreed that counterfeit bills of lesser-quality had been successfully passed off as real to the unsuspecting public. In light of this testimony, evidence that Rahm suffered from impaired perceptual ability would be highly probative as to the existence or absence of guilty knowledge.
[1415]*1415Rahm did not and could not contest the fact that the bills were counterfeit, that she possessed the bills, and that she attempted to use one of the bills to make a purchase. Rahm’s sole theory of defense was lack of knowledge. Knowledge was a “fact in issue” under Fed.R.Evid. 702. The proffered expert testimony addressed that issue specifically. If she had chosen to do so, Rahm herself could have testified to her lack of knowledge, but her testimony would not have related her atypical perceptual difficulties nearly as well as an expert’s conclusions from psychological test results.4 The proffered expert testimony had independent, significant probative value. Cf. United States v. Darby, 857 F.2d 623, 627 (9th Cir.1988) (proffered expert testimony meant to corroborate defendant’s testimony regarding lack of intent should have been admitted).
Although there is always a danger of undue prejudice with expert testimony “because of its aura of special reliability and trustworthiness,” Amaral, 488 F.2d at 1152, here that danger was minimal in comparison to the substantial probative value of Nelson’s proffered testimony. Neither are our other concerns under the fourth Amaral factor and under Fed.R.Evid. 403 — confusion of the issues, misleading the jury, undue delay, and needless presentation of cumulative evidence — at all implicated here. See Amaral, 488 F.2d at 1152 (listing countervailing considerations under fourth factor). Rahm’s brief two-day trial would hardly have been complicated unnecessarily or lengthened unduly by allowing the defendant to present her defense of lack of knowledge. Nelson’s proffered testimony satisfies the last Amaral factor.
We conclude that there was no justification for excluding the proffered expert testimony. The district court’s decision rested upon two incorrect legal conclusions and the testimony was otherwise admissible under the test established in Amaral.
C
We now determine whether the district court’s error was harmless. Under our test for nonconstitutional error, which we apply to errors ,as to the admissibility of expert testimony, we must reverse unless it is more probable than not that the error did not materially affect the verdict. See United States v. Echavarria-Olarte, 904 F.2d 1391, 1398 (9th Cir.1990).5 We conclude that the erroneous exclusion of Nelson’s testimony was not harmless.
In moving to exclude Nelson’s testimony, the government obviously was concerned about the testimony’s possible effect on the verdict. We think the government’s fears were justified. The circumstantial evidence of knowledge presented by the government was significant but not overwhelming, particularly in light of Rahm’s having remained in the store after having been informed of problems with the currency. Nelson’s testimony was highly probative of the defense Rahm sought to advance. It is true that Rahm might have been convicted even if Nelson’s testimony had been introduced, but we cannot say that that hypothetical outcome is “more probable than not.”
Our conclusion finds further support in the fact that the district court’s error deprived Rahm of her sole intended defense. Cf. United States v. Hartfield, 513 F.2d 254, 258 (9th Cir.1975) (absence of mental examination “could have seriously prejudiced” defendant because only available defense turned on mental condition). As discussed above, Rahm’s only possible defense was lack'of knowledge; she chose to present that defense through Nelson’s expert testimony regarding her perceptual difficulties. She had [1416]*1416a right to choose to present her defense exclusively through that testimony. The district court’s error was not harmless.6
IV
We conclude that the district court erroneously excluded Rahm’s proffered expert testimony. In requiring that expert psychological testimony relate to a mental “disorder,” and in requiring that the expert have a conclusive opinion, the district court imposed legally-erroneous standards. The expert testimony was admissible under the test established in Amaral. The error was not harmless. We reverse and remand for proceedings consistent with this opinion.
REVERSED AND REMANDED.