dissenting:
I respectfully dissent. The majority has produced an intellectually sluggish opinion, electing to sweep the district court’s legal (and factual) errors under the proverbial rug rather than undertake its responsibility to make sense of Texas’s jurisprudence regarding retardation, to apply it rigorously to Eric Moore’s case, and to provide guidance. Haphazardly-applied standards of review, casually-read caselaw, and superficially-scrutinized evidence make for an unfortunate combination; here, they result in shallow analysis and the wrong result. The only mitigation is that the majority opinion is unpublished, so it is not binding on anyone or any court.
I. Legal Background.
A. Standard of Review.
The majority’s errors run deep, beginning with its exaggeratedly deferential standard of review. Moore was required to show, by a preponderance of the evidence, that he satisfied the test for retardation established by Ex parte Briseno, 135 S.W.3d 1, 12 (Tex.Crim.App.2004), and its progeny.1 But though the district court acts as the finder of fact on the overall question of mental retardation,2 and though we review its factual findings for clear error,3 we apply a de novo review of legal questions as a necessary means of determining whether the ultimate factual question of retardation was decided without reversible error.
The majority fails to acknowledge it, but there is no doubt on that score: To the extent that errors of state or federal law “influence” the district court’s decision, even that court’s findings of fact are accorded no deference whatsoever.4 “[A] *75judgment based on a factual finding derived from an incorrect understanding of substantive law must be reversed,”5 because incorrect legal conclusions “taint” resulting factual findings.6 The clearlyeiToneous standard of review gives no shelter to a district court’s legal errors.
Similarly, the district court cannot immunize its findings from reversal merely by characterizing them as determinations of credibility. Anderson v. City of Bessemer City, 470 U.S. 564, 575, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985).7 Nor is the existence of some evidence to support the district court’s finding sufficient to rule out a determination of clear error entirely. Id. at 573, 105 S.Ct. 1504.
B. The Briseño Test.
After the Supreme Court decided Atkins v. Virginia, 536 U.S. 304, 122 S.Ct. 2242, 153 L.Ed.2d 335 (2002), leaving the details of substantive and procedural protections for mentally retarded convicts to the states, the Texas Court of Criminal Appeals (“TCCA”) adopted the general diagnostic framework of the American Association on Mental Retardation (“AAMR”)8 that is employed elsewhere in Texas state law. Briseño, 135 S.W.3d at 7. The Texas Legislature has taken no action pursuant to Briseño,9 and Texas courts have continued to apply it. The AAMR definition is therefore authoritative, at least to the extent that it was incorporated by the Briseño court, and subject to any modifications that Briseño requires.10
The Briseño definition of retardation has three parts that the defendant has the burden of proving by a preponderance of the evidence. Id. at 12. First, Briseño requires that putatively retarded individuals have “ ‘significantly subaverage’ general intellectual functioning,” “defined as an IQ of about 70 or below (approximately two standard deviations below the mean).”11 In practice, IQ tests typically are normed to an average IQ defined as 100, with a standard deviation of 15 or 16. AAMR 10th at 60-66 (discussing various IQ tests). Retarded persons are those *76with IQ’s two standard deviations or more below the mean, that is, in the bottom 2-3% of the population as measured by IQ testing.
Second, Briseño requires that low IQ be “accompanied by related limitations in adaptive functioning.” Briseño, 135 S.W.3d at 7 & n. 25 (internal quotation marks omitted). “Impairments in adaptive behavior are defined as significant limitations in an individual’s effectiveness in meeting the standards of maturation, learning, personal independence, and/or social responsibility that are expected for his or her age level and cultural group, as determined by clinical assessment and, usually, standardized scales.”12 Third, the defendant’s poor IQ and adaptive functioning deficits must have manifested themselves before the age of eighteen.13
II. Intellectual Functioning.
The district court concluded that Moore satisfied the first prong of the Briseño test by demonstrating an IQ of “about 70” or lower.14 The court explained that it had two bases for that finding: first, what the court took to be the state’s expert’s agreement with Moore that Moore’s IQ is impaired for purposes of Briseño, and second, Moore’s history of IQ testing.15 The court, contrary to the majority’s analysis, committed reversible error on both counts.
A. Dr. Mears’s “Agreement.”
Moore submitted the expert testimony of Dr. Anatolin Llórente, who administered an IQ test to Moore and testified that Moore’s IQ is below the Briseño threshold of 70. The state’s expert, Dr. Gary Mears, stated on direct examination that, on account of various scoring errors in, and necessary adjustments to, Moore’s IQ test scores, his IQ is significantly higher than 70, perhaps even over 80, and in any case not in the retarded range. 3 EH 239.16 Even before Mears identified flaws in Llorente’s testing procedure, he argued in his report that because Llorente’s test did not take “cultural factors and educational factors” into account, it critically understated Moore’s actual IQ and contributed to Llorente’s misdiagnosis.17 Mears’s conclusions, if adopted by the district court, would have precluded a finding of retardation.
The colloquy that opened Mears’s cross-examination, however, has proven a source of confusion for both the district court and *77the majority. The panel does not quote the entire relevant dialogue, so I will:
Q. Let’s start with IQ testing.
A. Yes, sir.
Q. Can we agree for the purposes of this hearing that Mr. Moore has significant limitations in intellectual functioning, the first prong of the mental retardation definition? That is my question.
A. I agree that he has limitations in intellectual functioning, yes.
Q. Now, do you agree to the extent we can move past that first prong of the mental retardation definition and focus on adaptive deficits?
A. Yes, sir.
Q. So you agree with Dr. Llórente that whether we use the DSM definition or the AAMR definition, the first prong is satisfied in your professional opinion?
A. The prong in terms of — I have some question [sic] about his accuracy of the score, but I would still, nevertheless, because of my opinion about the adaptive functioning, I will accept that.
Q. You do accept that?
A. I will accept it.
Q. Because I don’t want to spend half a day talking about it?
A. Right, I will accept it — regardless of the scoring errors. I still will accept it.
Q. Okay.
[Moore’s Attorney], Your Honor, I hate to rest on that, but I think I need to. I think we have moved past that prong.
THE COURT. All right. Proceed.
Q. I appreciate that, sir.
A. Thank you.
Q. You have helped us move along....
4 EH 3-4. The district court took this exchange as a concession by Mears and concluded that “both experts agree[d] that Moore satisfies” the Briseño IQ criterion. Moore v. Dretke, 2005 WL 1606437, at *5. The majority characterizes the testimony as “at best ambiguous” and defers to the district court’s view.
That is absurd; only one interpretation of Mears’s testimony is defensible, and it is not the district court’s or the majority’s. It makes no sense to suppose that Mears (1) submitted a report asserting that Moore had an IQ well in excess of the ceiling for a retardation diagnosis and attacking the accuracy of Llorente’s contrary evidence, (2) testified to that effect at great length on one day of a hearing (even bringing up additional concerns with Llo-rente’s evidence), and (3) then for no apparent reason conceded the issue entirely in the opening moments of his cross-examination, first thing the very next morning. That, however, is exactly what the district court and the majority say.
Neither Moore, nor the district court, nor the majority has posited an explanation for such a dramatic reversal. Nor can I imagine an explanation that survives a moment of sober reflection on the actual course of the district court proceedings. After eliminating rationalizations too fanciful even for Moore to suggest in his own defense (but which, somehow, satisfy the district court and the majority), what remains is only the conclusion that Mears made no concession at all.
A broader review of the record only bolsters that view. On direct examination, immediately after explaining his objections to the IQ test administered by Llórente, Mears emphasized that even if, hypothetically, Llorente’s IQ testing were accepted as accurate, he was “so convinced” that Moore’s adaptive functioning far exceeded the level of mentally retarded persons that whether Llorente’s IQ testing was adopted *78would not matter. 3 EH 220.18 Though the majority chooses to ignore it, that is almost precisely the view Mears repeated to Moore’s counsel on cross-examination. Mears’s testimony was consistent between direct and cross examination; there was no concession.
Plainly, Mears wished to present two independent bases for rejecting Moore’s Atkins claim: Because Mears felt that the adaptive-functioning prong was not even a close call in Moore’s case, it did not matter what the court decided on the IQ issue, so Mears did not, in the words of Moore’s counsel, “want to spend half a day talking about it.” Mears reserved his objections to Moore’s IQ scores, but all parties preferred to focus on Mears’s subsequent testimony regarding Moore’s adaptive functioning.19
This is not a case in which two plausible theories explain the evidence, nor is it a decision concerning a witness’s credibility where the panel must accord the district court heightened deference under the clearly-erroneous standard of review. See Anderson, 470 U.S. at 574-75, 105 S.Ct. 1504. The district court misunderstood the plain words of Mears’s testimony. The question is not whether Mears meant what he said, or whether what he said was believable or true; rather, the question is whether he said what he said, and that can be determined by the cold words in the transcript without reference to credibility, inflection, or demeanor.
Notwithstanding the majority’s headlong rush to defer to the district court, the “definite and firm conviction that a mistake has been committed” is inescapable. Id. at 573, 105 S.Ct. 1504. To the extent the court relied on Mears’s “agreement” in making its finding on Moore’s IQ, that decision “is implausible in the light of the record considered as a whole,” Rivera, 505 F.3d at 361, and the court committed clear error. That disposes of one of the district court’s two bases for finding that Moore’s IQ satisfies the requirements of Briseno.
B. Moore’s IQ.
1. Criteria.
The present panel’s next job, in a correct treatment of Moore’s case, would have been to identify the Texas state standard for deciding whether Moore’s IQ meets the requirements for retardation under Atkins. The basic question is whether Texas law treats the IQ of 70 spoken of in Brise-ño and AAMR 10th as a bright-line cutoff point. The majority studiously avoids thinking about that question. But it is impossible to evaluate the district court’s finding that Moore’s IQ is below the borderline set by Briseno without first determining what that borderline is, especially considering that the court’s factual findings would not be entitled to deference if the standard it applied were wrong. I will therefore fill that hole in the majority’s analysis.
*79The TCCA has left no doubt that an IQ below 70 (or at or below 70) is required for a finding of retardation under Atkins.20 In our post-Briseno decisions, this court has interpreted the Texas standard in the same way as the TCCA.21 In no case has a Texas court or the Fifth Circuit found mental retardation for Atkins purposes where its best guess as to IQ was above 70.
To be sure, IQ tests have a margin of error, and a person’s actual IQ may be below 70 even when an IQ test yields a higher score. Both the Briseno court and AAMR 10th speak of an IQ measurement in terms of probabilities and acknowledge that, considering the five-point margin of error typical in IQ tests, a diagnosis of retardation is sometimes possible even where a person’s IQ is measured to be above 70. See Briseño, 135 S.W.3d at 7 n. 24; AAMR 10th at 57.
Texas caselaw is explicit, though, in holding that an actual IQ between 70 and 75 is not sufficient for a finding of retardation.22 The Briseno petitioner presented scores of 72 and 74 and pointed out that, because of the margin of error, his true IQ could be lower. The TCCA rejected that argument: “[E]ven if a factfinder applied the statistical standard deviation, there is not enough evidence in this record that proves, by a preponderance of evidence, that applicant’s true IQ is lower than 72-74 rather than higher than 72-74.” Brise*80ño, 135 S.W.3d at 14 n. 53. If the Briseno court had meant to establish an IQ cutoff above 70, or not to establish a cutoff at all, or if it had envisioned simple application of an IQ test’s margin of error to the resulting score, it would have had no reason to make this observation; the scores of 72 and 74 would likely have been satisfactory by themselves as a matter of law. The Briseno court also expressly questioned the diagnostic usefulness of any margin of error where “the burden of proof is preponderance of the evidence, not a 95% confidence burden.” Briseño, 135 S.W.3d at 14 n. 53,23
The most accurate interpretation of Texas law, then, is that the lodestar of the Briseno analysis, as reflected in Briseno and its progeny, is an IQ of 70 or below (or, in accordance with some of the cited cases, an IQ below 70), regardless of the margin of error. Although IQ tests certainly have a margin of error, recognition of that margin merely permits evidence that a defendant’s true IQ is lower or higher than what was measured.24 Evidence may show that an IQ measurement between 70 and 75 is inflated, thus permitting a finding of “significantly subaverage general intellectual functioning,” just as the opposite may be shown regarding an IQ measured below 70, Briseño, 135 S.W.3d at 7 n. 24, and courts should be “flexible” in considering such evidence.25 But without such evidence, an IQ score above 70 is not itself indicative of retardation “merely because the low end of [a defendant’s] confidence band [is] below 70.” Clark, 457 F.3d at 446.26
*81Not only is the TCCA’s directive on this matter plain, but this court has adopted it. In Clark, the panel reasoned that “[t]he explanation provided by the [TCCA] in adopting its definition of mental retardation plainly foreclose[d]” the petitioner’s claim “that it was error for the state court [not] to find mental retardation where the lowest potential score in the confidence band of the 1983[IQ] test was below 70.” Clark, 457 F.3d at 446 (citing Briseño, 135 S.W.3d at 7 n. 24).
Moore was therefore required to prove that his IQ is below a bright-line cutoff of 70. Such a finding could be made in spite of, but not because of, IQ measurements above 70.
2. Moore’s IQ Scores.
Once the proper standard has been identified, the obvious next issue is whether the district court applied it correctly. Because the majority chooses not to identify that standard, it is unable to do anything much beyond shrugging its shoulders and babbling about clear error. The inescapable answer, however, is No, for reasons of both fact and law.
Moore’s IQ has been measured three times. When he was in first grade, his school administered a PMA test that showed his IQ to be 74. In 1991, the prison gave a WAIS-R test that yielded a score of 76. Lastly, Llórente administered a WAIS-III test in 2004 that produced a score of 66.27
All three of those test results were called into question at the evidentiary hearing. The PMA score, for example, is only a number; there is no evidence that it was properly scored or whether it was administered individually, as the test protocol requires, or to an entire school class. 3 EH 41-45. The vocabulary section of the WAIS-R, according to Llórente, was improperly scored, and in a way that may have slightly inflated the score.28 Llórente *82also testified concerning the “Flynn Effect,” the apparent increase in the average IQ scores in populations over time, as measured by a given IQ test. Because the WAIS-R was an older test when it was administered to Moore, Llórente suggested adjusting Moore’s score of 76 downward by about four points. 2 EH 175-78.29
Mears urged adjusting Moore’s score on the WAIS-R upward by as much as ten points to take Moore’s lack of education into account. 3 EH 239. Mears also stated that the decline in Moore’s measured IQ between the WAIS-R and the WAIS-III could be accounted for by Moore’s age, observing that though Moore’s verbal score stayed essentially constant, a decrease in his math score (in which abilities naturally decline with age) drove his second score down. 3 EH 246-49. Llórente made no attempt to correct for Moore’s age or educational level, despite the fact that Moore is well outside the demographic group to which the WAIS-III is normally applied. 3 EH 68-71, 228-30.
Mears also noted an error in Llorente’s scoring methodology. Llórente, it appears, administered vocabulary questions to Moore until Moore missed a total of six. According to Mears, however, the WAIS-III requires the administrator to ask questions until the subject misses six consecutive questions, and Moore had missed only four in a row when Llórente stopped. In effect, Llórente administered a partial test; we cannot know what would have happened to Moore’s score if Llórente had asked the next two questions. 3 EH 216-18.
Instead of grappling with those conflicting upward and downward adjustments, the district court gave the three scores equal weight, averaged them, reached an IQ of 72, applied the “five-point standard error of measurement,” and therefore concluded that Moore had borne his burden of proof. It is that finding, and the district court’s actual reasoning in making it, that the panel must consider, and yet the majority refuses to address it at anything resembling an acceptable level of detail.
S. Eirors.
To begin with, the district court’s IQ decision is not insulated by an evaluation of credibility. The court stated, in the context of its findings on Moore’s adaptive functioning, that it found Llórente more credible than Mears “[o]n all three criteria, but especially on Moore’s adaptive functioning[J” Moore v. Dretke, 2005 WL 1606437, at *13.30 In explaining its credi*83bility finding, the court made no mention of the experts’ IQ-related testimony but instead pointed to other aspects of their evaluation of Moore.
That finding, additionally, had no apparent effect on the district court’s reasoning with regard to Moore’s IQ. By averaging the IQ scores, the court declined to give additional weight to the test administered by Llórente. The court did not evaluate the upward and downward adjustments proposed by either expert and, in effect, treated them as a wash; it certainly did not say that it gave Llorente’s proposed adjustments any more credence than Mears’s.31 Though the district court certainly could have relied entirely on Llorente’s testimony and concluded that Moore’s IQ is 66, it chose not to. The court decided that Moore satisfied the IQ prong of its analysis not because it preferred Llorente’s testimony to Mears’s but because it mistakenly believed the experts to be in agreement. This panel therefore, in evaluating Moore’s IQ, need not give the court’s decision on this matter the special degree of deference that we accord findings involving credibility. Anderson, 470 U.S. at 575, 105 S.Ct. 1504.
The district court’s conclusion cannot survive a substantive review less slavishly deferential than the majority’s. For one thing, there is no legal or record support for taking an average of Moore’s IQ scores. Averaging IQ scores is, to say the least, a creative approach to their analysis and comparison and is highly unusual. Neither expert suggested, employed, or endorsed it.32 The district court assumed, without any evident backing, that averaging is a meaningful way to compare scores from different IQ testing protocols administered years apart and that the margin of error was the same for all three and was the same after the averaging as before. All of those assumptions are facially implausible, and the district court had no apparent reason to think any of them is correct.
Nor is there any precedent supporting such a move. When faced with a variety of IQ test results, courts instead (1) identify the single most reliable result,33 (2) narrow the results to a probable IQ range above or below 70,34 or (3) treat the IQ testing results as a wash that precludes a finding of retardation.35 All these approaches are preferable to just taking an average, especially considering that all three of the IQ scores are of doubtful reliability.36 The district court, in effect, made up its psychiatric testing methodolo*84gies as it went along, and the majority is compliant.
Granting, though, that the district court’s best guess as to Moore’s IQ was 72, the district court’s finding that Moore satisfied the IQ prong of Briseño collapses entirely. If Moore’s average IQ of 72 is his true IQ, then as a matter of law he is not retarded for purposes of Briseño and Atkins. Moreover, the fact that the district court found, based on the IQ evidence described, that Moore demonstrated “an IQ of ‘about 70 or below’ ” suggests that the district court failed to recognize Texas’s bright-line cutoff point correctly. As discussed above, Texas requires a finding of an IQ of 70 or below (or perhaps, in accordance with some cases, below 70). The district court committed reversible legal error if it operated on any other understanding of Texas law or reached a conclusion contrary to it.
If, on the other hand, the district court correctly understood the IQ cutoff point, the only way Moore’s average IQ could have justified a finding of retardation is if the court relied on extrinsic evidence as to Moore’s IQ that would have pointed to a lower result, for example, by showing that his scores were inflated or that the higher scores were entitled to less evidentiary weight. Cf. Rivera, 505 F.3d at 361-62. But the district court did no such thing. Instead, it seems just to have relied on the fact that IQ tests typically have a five-point margin of error. As explained above, if the court really did drag its best guess as to Moore’s IQ score below the 70-point threshold by applying the margin of error — and nothing in its reasoning suggests a more generous reading- — it did so contrary not only to Texas law but also to this circuit’s precedent.
That makes no difference to the majority, which refers to the district court’s “accounting for [the tests’] margins of error” without blinking, and evidently with a straight face. The majority is utterly silent, in fact, on the district court’s failure to apply the Briseño IQ cutoff point properly. Instead, it makes a last-ditch effort to salvage the district court’s decision by relying on the superficially comforting fact that, according to Llórente, Moore’s three IQ scores were “consistent.”
But the fact that three IQ scores, two above 70 and one below, may be “consistent” tells us nothing about whether Moore’s IQ is above or below 70, the question the court was bound to answer. After all, Mears testified that the scores were consistent as well, and the thrust of his testimony was the opposite of Llorente’s. The district court was still obligated to determine whether the high or the low end of the resulting range was most accurate, but it did not.37 The majority is without justification in ignoring that failure.
C. Summary Regarding IQ.
The district court applied an incorrect standard of law and misinterpreted the factual evidence in stating that Moore meets the IQ prong of the Briseño test. If Moore’s IQ is above 70, his claim for relief under Atkins must fail. While IQ tests have a margin of error, the law compels the district court to determine its effect; it makes no difference standing alone. Though the court may have been entitled merely to credit Llorente’s testimony and accept the IQ test he administered, it chose not to. The majority seeks to avoid *85reaching those conclusions by hiding behind the standard of review, but the majority cannot disguise its abdication of its responsibility to judge the district court’s findings.
The most charitable remedy for the district court’s errors is to remand for that court to hear Moore’s evidence again in light of the correct standards. That would require a new evidentiary hearing at which testimony, expert or otherwise, can be given and interpreted under a correct understanding of the law.
Such a forgiving outcome is by no means required, however. The panel is entirely justified in reversing the district court altogether, rendering judgment for the state, and submitting Moore to the punishment duly imposed on him. Moore had the burden all along of showing that he met the IQ prong of the Briseño test. Despite ample opportunity to present evidence, he failed to present a single complete IQ score on his own behalf. As I have explained, Llórente appears to have misadministered the WAIS-III in such a way that there is no telling how much higher Moore’s score would have been if correctly measured, and Texas courts and this court have repeatedly endorsed a hard denial of relief under such circumstances.38 For Moore to receive the relief he seeks, on the evidence he presented, is nothing short of a windfall.
None of the district court’s bases for finding in Moore’s favor on the first Brise-ño factor withstands scrutiny, nor do any of the majority’s justifications for affirming. All that remains is the majority’s craven refusal to confront and rectify the errors presented to it.
III. Adaptive Functioning.
Even if Moore’s IQ were sufficiently low under the Briseño test, he still has the burden to show the necessary deficits in adaptive functioning, the issue that (thanks partly to Mears’s generous cooperation with Moore’s counsel) took up the bulk of the Atkins hearing. Besides the respective experts, Moore relied primarily on the testimony of family members and acquaintances from his youth; the state, on Moore’s former teachers and prison guards. The court, after helpfully condensing the relevant evidence and testimony into a veritable biography of the petitioner, concluded that Moore suffered from significant deficits in adaptive functioning. Even giving due deference to the court’s determinations of fact and credibility, it erred as a matter of law.
A. Criteria.
Once again, the majority takes no interest in the criteria for reviewing the district court’s findings. But just as was the case with the IQ analysis, it is impossible to evaluate the district court’s opinion without understanding the adaptive functioning framework. Again, I explicate that standard and demonstrate the district court’s serious errors in applying it.
AAMR 10th discusses adaptive behavior in terms of “conceptual, social, and practical adaptive skills”39 and requires that an *86individual have significant deficits (placing him two standard deviations or more below normal) in at least one of those three categories. AAMR 10th at 73. But as explained in AAMR 10th at 81-82, the previous edition, which is AAMR 9th, employed a much different approach, dividing adaptive functioning into ten categories— communication, functional academics, self-direction, health and safety, social skills, leisure, self-care, home living, community use, and work — and requires significant deficits in two of those categories.
AAMR 9th was current at the time of Atkins.40 AAMR 10th was current at the time of Briseño, when the general AAMR framework was incorporated into Texas law.41 At first glance, then, it might seem most appropriate to require that adaptive functioning be analyzed under the AAMR 10th framework.42 But the Briseno court cited both editions'43 and never specified which adaptive-functioning test to use. In practice, other Texas courts, as well, have often referred to AAMR 9th.44 This court has typically followed whichever edition was used by the parties and experts.45 So have the district courts in this circuit.46 That approach is intuitively obvious, for it permits use of two comparably valid frameworks while guaranteeing that district courts will always have the necessary expert guidance in applying them.
In this case, confusingly, AAMR 9th and AAMR 10th were used inconsistently during Moore’s hearing. Mears’s report, though it never mentions either edition specifically, plainly follows the AAMR 9th standards. The categories of adaptive functioning he lists, and the two-category deficit requirement he mentions, are those of AAMR 9th. Res. Ex. 8 at 3. Mears’s direct examination, moreover, was structured around the AAMR 9th categories. 3 EH 250-57.
Llorente’s report and testimony are less explicit but inescapably lead to the same conclusion. After opining, in his report, that Moore had difficulties in his general academics and “vocational history,” Lló-rente segued into a discussion of additional fields of adaptive functioning — “social skills,” “daily living skills,” “other adaptive areas” — with the phrase “[a]s if these two *87factors were not enough.” Not only are these categories essentially those of AAMR 9th, but Llorente’s presentation demonstrates that he was looking for deficits in two or more specific areas of adaptive functioning, as AAMR 9th requires, rather than in one of AAMR lOth’s broad categories. Pet. Ex. 1 at 11.
Furthermore, when explaining adaptive functioning at the hearing, Llórente mentioned the categories of “social skills, communications, self-care, and those types of skills where an individual requires to be able to disengage his responsibilities in his society.” 2 EH 183. These are categories of the AAMR 9th framework and do not resemble the AAMR lOth’s three-part standard. At no time did Llórente specify a limitation in conceptual, social, or practical adaptive functioning; he consistently spoke of multiple, specific areas of adaptive functioning and said that Moore was limited in a variety of them.
Moore, however, argued to the district court that the AAMR 10th standard was applicable, 2 EH 17, and he entered excerpts from AAMR 10th into evidence, Pet. Ex. 7. Moore’s counsel, while cross-examining Mears, moved freely between the paradigms of the two editions.47 The court ultimately used the AAMR 10th three-part framework. For all its detailed reasoning and close examination of the evidence, the court thus failed to make its factual findings consistently with the standards laid out in the expert testimony.
A few examples will demonstrate how far adrift that failure led the district court. Though AAMR 10th gives some indication of how its standard relates to AAMR 9th’s categories and provides “representative skills” relevant to conceptual, social, and practical adaptive functioning, see AAMR 10th at 82 Table 5.2, its guidance is hardly self-explanatory. Llorente’s report, at its most definite point, finds that Moore has deficits in academics, social skills, vocational history, and daily living skills.
“Functional academics,” though, is only one of four AAMR 9th categories that the AAMR 10th table connects to conceptual adaptive skills, and it has no obvious counterpart among the AAMR 10th representative skills. “Social skills” is one of two AAMR 9th categories related to social adaptive skills, but there is no indication whether AAMR 10th and AAMR 9th use the terms identically. “Activities of daily living” is a representative skill area within practical adaptive skills in AAMR 10th (an area in which the court did not find significant limitations), but “self care,” “home living,” and “work” are three of the five AAMR 9th categories grouped under that new rubric. AAMR 10th gives no indication as to how deficits in AAMR 9th areas translate to the three-category framework.
At the same time, when the court considered the AAMR 10th representative skills (e.g., money concepts, obeying laws, avoiding victimization) — categories not mentioned in AAMR 9th — it did so essentially blind, not having heard experts explain those examples or how they fit into the all-embracing fields of adaptive functioning. One might put this still more starkly: No expert testified in support of the district court’s finding that Moore exhibits significant deficits in conceptual and social adaptive skills.
That is error that infects all the district court’s factual findings concerning *88adaptive functioning. Consider, for example, the court’s extensive treatment of Moore’s academic history, a subject much discussed by both experts. Those data, under AAMR 9th, might plausibly be pigeonholed under “functional academics,” as Llórente did in his report. But the nearest equivalent in AAMR 10th is “conceptual skills,” a category including such representative skills as “language” and “reading and writing.” AAMR 10th gives no indication — and the experts had no occasion to explain — how Moore’s grades in math, history, and science (for example) relate to his adaptive functioning under the framework the court employed.
The court also based its finding that Moore exhibits significant deficits in conceptual adaptive functioning largely on what it took to be his lifelong difficulty learning new skills (e.g., tying shoes, following directions, performing physical work in unison with others, riding a bicycle, driving a car).48 None of that has anything to do with the other two skills “representative” of conceptual adaptive functioning, “money concepts” and “self-direction.”
The district court’s use of the expert testimony in the context of AAMR 10th was therefore largely guesswork, uninformed by expert guidance. Because the witnesses spoke in terms of AAMR 9th, the categories of which map onto the AAMR 10th framework imperfectly, the court’s evaluation of Moore’s adaptive functioning is inherently flawed.
The majority does not even acknowledge this confusion, much less suggest any means of rectifying it. Plainly, though, no matter how deferential the standard or review may be, the difference between the two AAMR frameworks makes it impossible for the panel to evaluate whether the court properly applied evidence marshaled for AAMR 9th purposes to the AAMR 10th framework.49
The district court had no idea what it was doing in this regard. I cannot tell whether the court guessed right, and neither can the majority. The difference is that I find the prospect of the blind leading the blind in a death penalty case disturbing enough to avoid, while the majority cheerfully accepts it. Because of the absence of expert testimony, the review the majority applies is no review at all— little more than a rubber stamp. Whether the majority’s reticence flows from an exaggerated conception of the ability of generalist jurists to consider technical infor*89mation themselves, a stunted reliance on the standard of review, or mere inattention, I do not know. Regardless, it is obvious that what the panel ought to have done (if it reached Moore’s adaptive functioning at all) is decline to review the district court’s findings and remand for evaluation of the evidence consistently with the expert testimony that may be elicited in a further hearing.
B. The Additional Briseno Factors.
The Briseno court, 135 S.W.3d at 7 n. 25, 8, acknowledged that the criteria for adaptive functioning are “exceedingly subjective” and encouraged the use of standardized tests for their evaluation. The court, id. at 8, also suggested a number of “other evidentiary factors which factfin-ders in the criminal trial context might also focus upon in weighing evidence as indicative of mental retardation or of a personality disorder”:
• Did those who knew the person best during the developmental stage — his family, friends, teachers, employers, authorities — think he was mentally retarded at that time, and, if so, act in accordance with that determination?
• Has the person formulated plans and carried them through or is his conduct impulsive?
• Does his conduct show leadership or does it show that he is led around by others?
• Is his conduct in response to external stimuli rational and appropriate, regardless of whether it is socially acceptable?
• Does he respond coherently, rationally, and on point to oral or written questions or do his responses wander from subject to subject?
• Can the person hide facts or lie effectively in his own or others’ interests?
• Putting aside any heinousness or gruesomeness surrounding the capital offense, did the commission of that offense require forethought, planning, and complex execution of purpose?
Id. at 8-9.50
The district court acknowledged the existence of the Briseno factors but refused to consider their substantive application. Explaining that decision, the court interpreted Briseno as saying that the factors would apply only in those Atkins cases in which a court must decide whether a defendant is retarded or has a personality disorder. Because it did not understand the state to have argued, or presented any evidence, that Moore had such a disorder, the court considered the factors to be legally and factually irrelevant. It also said that it did not have to consider any factors not part of the AAMR test, and, lastly, it relied on the factors’ being discretionary. For these reasons, it chose not to take any of the Briseño factors into account. Moore v. Dretke, 2005 WL 1606437, at *5 n. 6.
The majority happily endorses the district court’s conclusion, failing to recognize that none of those reasons is an adequate ground for refusing to consider the Briseño factors. All, in fact, contain errors of law or clearly erroneous findings of fact,51 or at the very least indicate the court’s *90refusal to consider “relevant factor[s] that should have been given significant weight.”52 The court’s choice to disregard the Briseno factors is an abuse of discretion almost by definition.
To begin with, the fact that the AAMR does not mention those seven factors does not exclude their applicability in Texas law; the Briseno test supersedes the AAMR and is the law that controls here.53 The TCCA can modify the AAMR definition at its discretion, and in Briseno it did so. The district court was wrong as a matter of law to ignore Texas state courts in this way.
The court also erred as a matter of law in saying that those factors are inapplicable where personality disorders are not at issue. Briseno did involve a defendant who claimed to be retarded and who the state said merely suffered from a personality disorder. Other courts have used the Briseno factors to make a similar distinction.54 But the TCCA has understood or described the Briseno factors as being applicable in all cases,55 and so has this court.56 Whether the state argued or presented evidence related to personality disorders is therefore irrelevant to the factors’ applicability.
The district court’s reliance on a legally erroneous understanding of the Briseno factors’ scope of application constitutes an abuse of discretion and is reversible. Even if, “arguendo (and dubitante),”57 Briseno did limit the application of those factors to cases in which a personality disorder is presented as an alternative to a diagnosis of retardation, the district court was plainly incorrect in saying that the state had not argued or presented evidence that Moore’s adaptive functioning deficits may be caused by such a disorder. On both direct and cross-examination, Mears consistently suggested an “antisocial” personality as an alternative explanation to mental retardation.58 That is es*91sentially identical to the circumstances of Briseno, in which “the State’s expert found no mental retardation but did find evidence consistent with antisocial personality disorder.” Briseño, 135 S.W.3d at 13.
More importantly still, the district court’s statement that “there is no evidence to support a finding that a personality disorder is responsible for Moore’s cognitive and adaptive deficits,” Moore v. Dretke, 2005 WL 1606437, at *5, is belied by the court’s own analysis of Moore’s social adaptive functioning, which deals extensively with Moore’s disruptive behavior in school, social isolation, violence, and strained family relationships. Considering that both the state and the court’s factual findings invited consideration of whether Moore has an antisocial personality, the district court should have considered the factors set out in Briseno for just that purpose. In short, the court’s rejection of the Briseno factors was “based on an erroneous view” of the law — its impression that the factors are not applicable at all— and also on a “clearly erroneous assessment of the evidence” — its failure to recognize relevant evidence at the hearing.
After one rejects the district court’s first reasons for ignoring the Briseno factors, all that remains is the court’s naked reliance on the factors’ being discretionary. That, standing by itself, is not an acceptable basis for refusing to consider them. Though the TCCA has described the factors in discretionary terms, there is no Texas case in the years since Briseno in which the court, in the way the district court acted here, explicitly refused to consider the factors.
In fact, a review of post -Briseno cases shows that where Texas trial judges rather than juries make the Briseno analysis, they consistently — perhaps invariably— consider some or all of these factors, and the TCCA, after Briseno, has used almost mandatory language in describing them.59 This court, moreover, has referred to the factors as “correct definitions of mental retardation” for purposes of the governing Texas law. Moreno v. Dretke, 450 F.3d 158, 164 (5th Cir.2006). Evidence relevant to the factors was elicited in both the state trial and the federal evidentiary hearing, and all seven appear at least relevant to Moore’s case.60 Considering the factors’ *92obvious centrality to the Briseno test and relevance to this case, together with the absence of any justification for ignoring them, the district court’s choice was mere caprice — in other words, an abuse of discretion.
The district court had no legally cognizable reason for ignoring the Briseno factors, and the majority appears to recognize the precariousness of the district court’s refusal to address the evidence as the TCCA intended. The majority’s effort to justify that refusal, however, is nothing short of sophistic.
The majority completely disregards the district court’s mistaken legal understanding of the factors’ applicability. The majority does recognize the volume of evidence before the district court that was related to the factors, but the majority does not seem to understand how badly that cuts against its conclusion.61 If there was so much evidence relevant to the factors, after all, the district court’s reliance on the supposed lack of relevant evidence looks especially weak. Instead, the majority brazenly pretends that the district court considered the factors “implicitly.”62
That is pure fiction. For all its elaborate deference to the district court elsewhere in its shaky opinion, the majority fails to take the district court at its word in one of its most unambiguous statements, namely: “This Court will not analyze these factors for four reasons[.]” The only acceptable reading of that language is that the district court did not analyze those factors. Because all of the district court’s reasons are legally inadequate, there is no sense at all in the majority’s concocting imaginary factual findings to which to defer.
The great irony in the majority’s treatment of the evidence relevant to the Brise-ño factors is that though the majority happily endorses the district court’s application of that evidence to technical psychiatric criteria that it did not understand, the majority refuses to censure the district court for failing to apply the evidence to criteria, created for judicial use, that are critical under state law. The majority has elected to rewrite an indefensible district court decision rather than review it on its own merits.
IV. Conclusion.
In light of the record and the applicable law, the district court committed both legal and factual reversible errors in finding that Moore is retarded, for Atkins purposes, under Texas law. A clear-headed evaluation of the evidence in light of the correct legal standards undeniably poses a more difficult obstacle for Moore than the one he actually faced, considering his history of IQ testing and the weight of the Briseno factors. Because of the majority’s *93evident indifference to state legal standards, however, that evaluation will never come.
Rather than correct the district court’s undeniable errors, the panel majority washes its hands of Moore’s case under the guise of respecting the clearly-erroneous standard of review. That standard, however, cannot possibly justify the majority’s limpness in the face of those mistakes. The effect, which must be attributed entirely to the majority itself, is to abandon any attempt to get the right result in a complex case, to leave a difficult (and still embryonic) patch of constitutional law unimproved, and to mock, on the flimsiest of foundations, the honest efforts of Texas state courts to apply Atkins.
The only redeeming feature of the majority’s misguided opinion is that, as I noted at the outset, it is unpublished and therefore binding as precedent on no one except the parties to this case as it affects only this case. The correct resolution of these important questions, as they relate to the implementation of Atkins in Texas and in this circuit, must await another day and another panel. Moreover, because, under Atkins, the states are left to fashion their own response to that decision, the majority’s mangling of Texas legal standards is subject to ready correction by the TCCA or the legislature.
I respectfully dissent.
. Clark v. Quarterman, 457 F.3d 441, 444 (5th Cir.2006) (Dennis, J.); see also Morris v. Dret-ke, 413 F.3d 484, 497 (5th Cir.2005).
. Clark, 457 F.3d at 444.
. Rivera v. Quarterman, 505 F.3d 349, 361 (5th Cir.2007).
. G.M. Trading Corp. v. Comm'r, 121 F.3d 977, 980 (5th Cir.1997). Nor is the panel permitted to defer to the district court in its *75interpretation of Texas's definition of mental retardation. See Hulin v. Fibreboard Corp., 178 F.3d 316, 318 (5th Cir.1999) (Dennis, J.) (citing Salve Regina Coll. v. Russell, 499 U.S. 225, 231, 111 S.Ct. 1217, 113 L.Ed.2d 190 (1991) ("We conclude that a court of appeals should review de novo a district court’s determination of state law.”))
. Mobil Exploration & Producing U.S., Inc. v. Cajun Constr. Servs., Inc., 45 F.3d 96, 99 (5th Cir.1995).
. Id. at 99 n. 6 (quoting Fulton Nat'l Bank v. Tate, 363 F.2d 562, 567 (5th Cir.1966)).
. The Anderson Court explained,
Documents or objective evidence may contradict the witness' story; or the story itself may be so internally inconsistent or implausible on its face that a reasonable factfinder would not credit it. Where such factors are present, the court of appeals may well find clear error even in a finding purportedly based on a credibility determination.
Anderson, 470 U.S. at 575, 105 S.Ct. 1504.
. See AAMR, Mental Retardation. Definition, Classification, and Systems of Support (10th ed. 2002) ("AAMR 10th”).
. Neal v. State, 256 S.W.3d 264, 270 (Tex. Crim.App.2008) (“Over four years later, the Texas legislature still has not enacted any legislation on this matter.”), cert. denied, - U.S. -, 129 S.Ct. 1037, 173 L.Ed.2d 471 (2009).
. See Briseño, 135 S.W.3d at 9 & n. 30 (noting that the legal definition of retardation is different from, though informed by, "psychological diagnostic criteria”).
. Id. at 7 & n. 24 (quoting American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders 39 (Text Revision, 4th ed. 2000) ("DSM-IV”)). See also AAMR 10th at 35, 57 (discussing the AAMR 10th threshold IQ score for retardation).
. Briseño, 135 S.W.3d at 7 n. 25 (quoting American Association on Mental Deficiency, Classification in Mental Retardation (Grossman ed. 1983)). See also AAMR 10th at 41 (defining "adaptive skills" as those abilities "that have been learned by people in order to function in their everyday lives”).
. See Briseño, 135 S.W.3d at 7; AAMR 10th at 1.
. Moore v. Dretke, No. 603-CV-224, 2005 WL 1606437, at *5 (E.D.Tex. Jul. 1, 2005) (citing AAMR, Mental Retardation: Definition, Classification, and Systems of Support (9th ed. 1992) (“AAMR 9th”)).
. The court stated, in full,
With a five-point standard error of measurement and both experts agreeing that Moore satisfies this criterion for mental retardation, the Court concludes that Moore has proven by a preponderance of the evidence that he satisfies the AAMR criterion of subaverage intellectual functioning defined as an IQ of "about 70 or below."
Moore v. Dretke, 2005 WL 1606437, at *5.
. I cite the evidentiary hearing transcript in the form "[volume] EH [page].”
. Mears was not able to administer an IQ test himself, because Llórente had given one only three months before, 3 EH 216, and a six-month interval is needed for accuracy. Hall v. State, 160 S.W.3d 24, 30 n. 14 (Tex.Crim.App.2004).
. The majority lamely counters that on direct examination Mears "never clearly states that the scores [qualifying Moore on the IQ prong] are invalid.” That is not only irrelevant, but a flat misrepresentation of what happened at the hearing. The state did not need to show that Llorente's test was “invalid,” only that it was inaccurate or unreliable, and that is exactly the opinion Mears presented. Relying on Mears's failure use the word "invalid” in describing Llorente's test when he actually did testify that it was incorrectly administered and subject to a variety of upward adjustments is no way for the majority to treat a transcript that, read sensibly, provides no support for its opinion.
. The majority, in effect, penalizes the state because its expert cooperated with the court — to the evident gratitude of Moore's counsel — in expediting the hearing by avoiding what the parties perceived as unnecessary discussion.
. See, e.g., Williams v. State, 270 S.W.3d 112, 132 (Tex.Crim.App.2008) (distinguishing between the defendant's "70 and below IQ scores" and his "70 and above IQ scores”); Hunter v. State, 243 S.W.3d 664, 669-71 (Tex.Crim.App.2007) (noting defense expert’s testimony that 70 "is considered the ‘cutoff point’ "), cert. denied, - U.S. -, 129 S.Ct. 51, 172 L.Ed.2d 54 (2008); Howard v. State, 153 S.W.3d 382, 386 (Tex.Crim.App.2004) (" 'Significantly subaverage general intellectual functioning' is defined as an IQ of 70 or below.”); see also Neal, 256 S.W.3d at 272-73 ("significant sub-average general intellectual functioning, usually evidenced by an IQ score below 70”); Ex parte Blue, 230 S.W.3d 151, 163 (Tex.Crim.App.2007) (same). The fact that Texas courts have referred to 70 as a "presumptively retarded score,” see, e.g., Ex parte Taylor, No. WR-48498-02, 2006 WL 234854, at *3 (Tex.Crim.App.2006) (per curiam) (unpublished) (Johnson, J. concurring), supports this interpretation.
I acknowledge an otherwise-unsupported statement in Ex parte Modden, 147 S.W.3d 293, 298 (Tex.Crim.App.2004), that a "70-75 score” "generally indicates subaverage general intellectual functioning.” That assertion, however, not only is dictum — considering that the defendant's IQ had been measured, in two tests, at 58 and 64 — but also is contrary to the overwhelming weight of TCCA authority; the statement, to my knowledge, has not been repeated in any later decision.
. See Taylor v. Quarterman, 498 F.3d 306, 308 (5th Cir.2007) ("[T]he only IQ test taken of Taylor prior to his turning eighteen yielded a result of 75, above the mild retardation cut off of 70.”), cert. denied, - U.S. -, 128 S.Ct. 1739, 170 L.Ed.2d 543 (2008); Woods v. Quarterman, 493 F.3d 580, 586 (5th Cir.2007) ("[T]he court noted ... the existence of four IQ test scores placing Woods above the seventy-point cutoff."); In re Mathis, 483 F.3d 395, 397 (5th Cir.2007) ("Typically, a person’s IQ must be measured at 70 or below to qualify as mentally retarded.”); Clark, 457 F.3d at 445 (stating that IQ of 70 is "the rough cut-off for mental retardation”); In re Salazar, 443 F.3d 430, 433 n. 1 (5th Cir.2006) (per curiam) (treating 70 as the “cutoff score” even where an IQ score could have been "adjusted” to below 75). See also Wilson v. Quarterman, No. 6:06-CV-140, 2009 WL 900807, at *5 (E.D.Tex. Mar. 31, 2009) (quoting Texas state trial court findings stating that 70 is "considered the border below which a person is mildly mentally retarded”).
. See Neal, 256 S.W.3d at 273-75 (deferring to jury's finding of no preponderance based on test scores of 70, 72, and 87); Taylor, 2006 WL 234854, at *6 (Johnson, J., concurring) (no preponderance based on score of 75); Ex parte Rodriguez, 164 S.W.3d 400, 403, 405 (Tex.Crim.App.2005) (Cochran, J., concurring) (score of 71, according to defense expert, "not in the retarded range”; 70 a "cutoff level”).
. That makes excellent statistical sense. By definition, even taking the margin of error into account, an average IQ score above 70 cannot by itself prove an IQ of 70 or below by a preponderance of the evidence, for when the center of a confidence interval is higher than 70, the probability that the score is below 70 cannot be above 50 percent. The probability that a subject’s true IQ lies above 70 is necessarily above 50 percent; in the absence of other data, that precludes finding an IQ below 70.
. See Briseno, 135 S.W.3d at 14 n. 53; see also Hunter, 243 S.W.3d at 669-70; Taylor, 498 F.3d at 308 ("The administrator of the test thought Taylor was capable of performing better than 75. While Taylor's expert concluded that this test result overstated Taylor's IQ by seven points, the trial court was not unreasonable in finding otherwise.”). In Bri-seño, the trial court determined that the most accurate measurements of IQ yielded results of 72 and 74. The TCCA quoted and adopted the trial court’s finding that "[t]he preponderance of the evidence does not show that these test scores over-state the actual intellectual functioning of Applicant; the evidence in fact showed that there are good indications that the test scores understated Applicant’s intellectual functioning.” Briseno, 135 S.W.3d at 14.
The decision in Briseno therefore reflects a combination of (1) skepticism toward any given IQ measurement and (2) searching for a true IQ measurement at (or maybe below) the threshold of 70. This is consistent with the treatment in AAMR 10th, at 57-59, of variations in IQ measurements and fully explains why some Texas courts, see, e.g., Neal, 256 S.W.3d at 272-73, state that an IQ score below 70 "usually” satisfies the IQ prong. One way or another, evidentiary flexibility in determining whether a standard is met is not to be confused with the absence of any standard at all. See Clark, 457 F.3d at 446 (explaining that Texas courts "make a flexible determination” of whether a defendant's IQ is above or below 70).
. See Clark, 457 F.3d at 444-46. See also Williams v. Quarterman, 293 Fed.Appx. 298, 309-10 (5th Cir.2008) (per curiam) (finding an IQ above 70 where petitioner had tested at 70 or 71 repeatedly, and the district court suspected malingering); In re Hearn, 418 F.3d 444, 447 & n. 4 (5th Cir.2005) (considering measurement error in granting permission to file a successive habeas petition based on IQ of 74).
. One district judge from the court a quo has presented a compelling justification for a bright-line rule.
*81Even though IQ testing and evaluation of adaptive limitations involves ranges, the professional organizations and the courts have adopted standards or criteria, which establish a cut-off point. No matter which criteria are chosen, there will be some individuals who are very close to the border of the cutoff — even if only slightly higher. Saying that a Defendant should be classified as mentally retarded simply because he or she is "borderline” would make the cutoff, which is already based on an evaluation of ranges, impossibly vague. There will always be a Defendant who is close to, but just above, the level of the last person found to be mentally retarded who could therefore be termed "borderline.” That Defendant would then be in a new "borderline” subcategory of the classification of mentally retarded. The concept of "borderline” would swallow any definition or standard.
Simpson v. Quarterman, 593 F.Supp.2d 922, 941 (E.D.Tex.2009).
. The majority notes that Llórente also administered a TONI-2 test that showed an IQ of 60, but even Llórente seems not to have considered that test to be a measure of general intelligence. His conclusion reflects the standard professional view of the TONI-2. See, e.g., Hall v. Quarterman, 534 F.3d 365, 376 (5th Cir.2008). In any event, though the district court mentioned that score, it did not rely on it. See Moore v. Dretke, 2005 WL 1606437, at *4-*5.
. The WAIS-R verbal section includes a series of progressively more difficult vocabulary questions. Test subjects are asked to define the fourth word in a list, with full credit assumed on the first three. But if the subject receives no credit on any one of questions 4 through 8, the administrator is required to ask and score questions 1 through 3. Petitioner Exhibit ("Pet.Ex.") 13. Moore received no credit on question 7, "fabric” (he evidently answered “clothes”), but the administrator did not score questions 1 through 3, "bed,” "ship,” and "penny,” to confirm that Moore knew the words. Respondent Exhibit ("Res. Ex.”) 7 at 33; 2 EH 171-74. This may have inflated the score, perhaps by "one or two points.” 2 EH 174. Moore answered "ship” and "penny” correctly on the WAIS-III test *82that Llórente administered in 2004, but Lló-rente said one cannot assume that Moore would have known the terms in 1991. 3 EH 52-55.
. The Flynn Effect “has not been accepted in this Circuit as scientifically valid.” Mathis, 483 F.3d at 398 n. 1 (citing Salazar, 443 F.3d at 433 n. 1). More importantly for our purposes, it is not evident whether Texas courts would permit consideration of the Flynn Effect. Neal, 256 S.W.3d at 273 ("We have previously refrained from applying the Flynn Effect, noting that it is an 'unexamined scientific concept' that does not provide a reliable basis for concluding that an appellant has significant sub-average general intellectual functioning.”) (quoting Blue, 230 S.W.3d at 166 ("This Court has never specifically addressed the scientific validity of the Flynn Effect. Nor will we attempt to do so now.”)). Even if both of Llorente's adjustments to the WAIS-R were adopted, the test would still yield a measurement of 72, above the Texas threshold.
. It is puzzling that the court could find one expert more credible than the other on the third element — onset before age eighteen— because both parties and, as the district court noted, Moore v. Dretke, 2005 WL 1606437, at. *15, both experts agreed that that element of the retardation test was met. Apparently the district court was intent on painting with a *83broad brush the fact that it liked Llorente’s testimony better than Mears’s.
. The district court did discount the state's argument that Moore was malingering on the test administered by Llórente, but Mears (contrary to the statement in the court’s memorandum opinion) did not express that view himself, so that conclusion did not entail a credibility assessment as between the two experts.
. The majority’s statement to the contrary has no basis in the record. Reading support for averaging IQ scores into Llorente's testimony about the "consistency” of Moore’s three IQ scores, as the majority tries to do, merely distorts the hearing testimony transcript with wishful thinking.
. See Rivera, 505 F.3d at 361-62; Taylor, 2006 WL 234854, at *6 (Johnson, J., concurring).
. See Briseño, 135 S.W.3d at 14 & n. 53; see also Clark, 457 F.3d at 445-46.
. See Blue, 230 S.W.3d at 165-66 (stating that where the defendant suggests adjusting a "partial” score, above 70, to below 70, "we will simply regard the record as ... devoid of any reliable IQ score”); Lewis v. Quarterman, 541 F.3d 280, 284 (5th Cir.2008); see also Howard, 153 S.W.3d at 387.
. Under such circumstances, even Moore's counsel, at oral argument, conceded that the results must be "weighed.”
. Even if, viewing its words most loosely, it decided that Moore's IQ was at the low end of the confidence interval based on the experts' supposed agreement that Moore had satisfied the first Briseño prong, the court's conclusion would still fail, because, as explained above, there was no such agreement.
. See Blue, 230 S.W.3d at 165-66 (stating that where the defendant suggests adjusting a "partial” score, above 70, to below 70, "we will simply regard the record as ... devoid of any reliable IQ score”); Lewis v. Quarterman, 541 F.3d 280, 284 (5th Cir.2008); see also Howard, 153 S.W.3d at 387. See also Williams, 293 Fed.Appx. at 311 ("Because Williams carries the burden of proving mental retardation, failure to prove the first Briseno prong should end our inquiry.”).
. AAMR 10th provides examples of "representative skills” within these categories. Under conceptual skills, the AAMR lists "language (receptive and expressive),” "reading and writing,” "money concepts,” and "self-*86direction;” under social skills, "interpersonal,” "responsibility,” "self-esteem,” "gullibility (likelihood of being tricked or manipulated)," "naivete,” "follows rules,” "obeys laws,” and "avoids victimization;" under practical skills, "activities of daily living,” "instrumental activities of daily living,” "occupational skills,” and "maintains safe environment.” AAMR 10th at 42.
. See Atkins, 536 U.S. at 308 n. 3, 122 S.Ct. 2242.
. See Briseño, 135 S.W.3d at 8 n. 29.
. See Ex parte Chester, No. AP-75037, 2007 WL 602607, at *1 & n. 9 (Tex.Crim.App. Feb. 28, 2007) (unpublished) (interpreting Briseño as requiring "significant deficits in adaptive functioning, usually expressed by limited conceptual, social, and adaptive skills”).
. See generally Briseño, 135 S.W.3d at 7-8.
. See Modden, 147 S.W.3d at 295-96 & n. 9 (quoting and attributing to Briseño the AAMR 9th standard).
. See Williams, 293 Fed.Appx. at 311-12 (AAMR 9th); Rivera, 505 F.3d at 357 (same); Woods, 493 F.3d at 585 (same); Morris, 413 F.3d at 487-88 & n. 1 (AAMR 10th).
. See Simpson, 593 F.Supp.2d at 937-38; Rivera v. Dretke, No. B-03-139, 2006 WL 870927, at *10 (S.D.Tex. Mar. 31, 2006) ("[Tjhis Court will utilize the 1992 AAMR criteria [i.e., AAMR 9th] ... because it is the definition that has been endorsed by prior courts and because it is the definition upon which both sides agree, despite the fact that Petitioner discusses both definitions without advancing one over the other.”) (same), aff'd in part, vacated in part, 505 F.3d 349 (5th Cir.2007), cert. denied, — U.S. -, 129 S.Ct. 176, 172 L.Ed.2d 44 (2008).
. 4 EH 44-47 (questioning Mears on Moore’s "self-esteem,” "gullibility," obedience to law, "ability to avoid victimization,” and "occupational skills,” all listed in AAMR 10th as "skills” relevant to social adaptive functioning); 4 EH 52-56 (questioning Mears on Moore’s "academic functioning”); 4 EH 60 ("I have got before you ... the three broad categories from the AAMR. I am going to start and look at conceptual.”).
. The court counted many of these difficulties a second time in the context of Moore's practical adaptive functioning, as to which the court did not find a significant deficit.
. In Williams, the TCCA reviewed and rejected a defendant's appeal from a jury verdict rejecting his Atkins claim in which, although an expert referred to the AAMR 9th categories, the jury was instructed on the AAMR 10th categories. In that case, however, the TCCA opined that the jury could have based its conclusions on grounds other than adaptive functioning — for example, the defendant's IQ — and in any case the standard of review applied to the verdict was far more deferential than is the clear error review here. Williams, 270 S.W.3d at 132 (evaluating whether the verdict was "so against the great weight and preponderance of the evidence as to be manifestly unjust”).
In Gallo v. State, 239 S.W.3d 757, 778 (Tex.Crim.App.2007), cert. denied, - U.S. -, 128 S.Ct. 2872, 171 L.Ed.2d 813 (2008), the TCCA permitted a jury instruction defining "adaptive behavior" as "the effectiveness with or degree to which a person meets the standards of personal independence and social responsibility expected of the person’s age and cultural group” instead of enumerating the categories. (The experts had used the AAMR 9th approach.) But that is merely the definition found in § 591.003 of the Texas Health and Safety Code, which the Briseño court largely incorporated. Briseño, 135 S.W.3d at 7 n. 25.
. The Briseno court evidently expected that these factors, being intuitively simple and nontechnical, would be helpful to courts in sorting out contradictory expert testimony concerning the “exceedingly subjective” criteria for adaptive functioning in close cases. Btiseño, 135 S.W.3d at 8.
. Ratliff v. Stewart, 508 F.3d 225, 229 (5th Cir.2007) (internal quotation marks omitted); see also Unger v. Amedisys Inc., 401 F.3d 316, 320 (5th Cir.2005); Chaves v. M/V Medina Star, 47 F.3d 153, 156 (5th Cir.1995).
. In re Volkswagen of Am., Inc., 545 F.3d 304, 310 n. 4 (5th Cir.2008) (en banc) (quoting Kern v. TXO Prod. Corp., 738 F.2d 968, 970 (8th Cir.1984)), cert. denied, - U.S. -, 129 S.Ct. 1336, 173 L.Ed.2d 587 (2009). See also Dell Computer Corp. v. Rodriguez, 390 F.3d 377, 385 n. 14 (5th Cir.2004) (“Thus, 'when judicial action is taken in a discretionary matter,' that action may be set aside by a reviewing court if 'it has a definite and firm conviction that the court below committed a clear error of judgment in the conclusion it reached upon a weighing of the relevant factors.' ") (quoting United States v. Walker, 772 F.2d 1172, 1176 n. 9 (5th Cir.1985)).
. See Modden, 147 S.W.3d at 299-300 (“[Tjhis Court in Ex parte Briseno adopted the [AAMR] criteria ... and some other evidentia-ry factors in addressing Atkins mental retardation claims.”) (citation omitted) (Henry, J., dissenting); see also Mathis, 483 F.3d at 397 (stating that retardation depends on the AAMR definition "and associated factors" established by Briseno ); Clark, 457 F.3d at 445 (“Although the Court [in Atkins ] did refer to the clinical definitions of mental retardation promulgated by the AAMR ... it did not dictate that the approach and the analysis of the State inquiry must track the approach of the AAMR ... exactly.”).
. See, e.g., Williams, 293 Fed.Appx. at 312.
. See, e.g., Hunter, 243 S.W.3d at 666-67; Gallo, 239 S.W.3d at 769-70; Modden, 147 S.W.3d at 296 & n. 12.
. See, e.g., Rosales v. Quarterman, 291 Fed. Appx. 558, 562 (5th Cir.2008) (per curiam), cert. denied, — U.S. -, 129 S.Ct. 1317, 173 L.Ed.2d 596 (2009) ("In Briseno, the TCCA also laid out a number of guidelines that courts can consider when making mental retardation determinations.”).
. Puckett v. United States,-U.S.-, 129 S.Ct. 1423, 1431, 173 L.Ed.2d 266 (2009).
. 3 EH 223 (stating that Moore's adaptive problems are caused by antisocial tendencies, not adaptive problems in the retardation sense); 3 EH 252 (explaining that Moore is not retarded but rather is "more on the antisocial spectrum”); 4 EH 41 (distinguishing *91between adaptive functioning for purposes of retardation and diagnosis of antisocial personality); 4 EH 55-56 ("I think he has does have some learning disabilities [as distinguished from retardation].... [The significance of a child’s academic problems] depends on if the child is antisocial.... He may not have been motivated.”); 4 EH 63-64 (state’s attorney questioning Mears on the difference between “problem” or “antisocial” behavior and adaptive deficit; poor performance in school not an adaptation issue if it derives from antisocial tendencies or poor motivation).
. See Modden, 147 S.W.3d at 297 (”[T]he trial court did not use the Briseno factors in its findings because it did not have the benefit of our order in that case. Nonetheless, we conclude that the trial court’s findings are supported by the record.”); Ex parte Elizalde, No. WR-48957-02, 2006 WL 235036, at *3 (Tex.Crim.App. Jan. 30, 2006) (unpublished) (Johnson, J., concurring) ("In Briseno, this Court examined seven factors to gauge the level of an individual’s adaptive functioning. However, applicant provides no evidence ... to address any of the factors outlined in Brise-ño."). See also Gallo, 239 S.W.3d at 777 (“The ultimate issue of whether a defendant is mentally retarded for purposes of the Eighth Amendment ban on excessive punishment is one for the finder of fact, based upon all of the evidence and determinations of credibility.”).
. In fact, the district court and the majority resolutely ignore the unmistakable fact that all seven of the Briseno factors likely weigh against Moore. The adults who knew him when he was a child were fairly unanimous that while they regarded him as mentally sub-average, they never thought he was retarded. *92Moore's role in his crime — including the ruse it involved and the manner in which Moore duped his co-defendants out of the proceeds and tried to shift responsibility for the victim's death — undeniably indicates a capacity for planning, leadership, and subterfuge. There was no indication that his conduct has ever been irrational. He answers questions directly and coherently, as evidenced by his interviews with Llórente, Mears, and the police after his arrest; no testimony suggested otherwise. Merely to affirm the district court bespeaks a refusal to confront that contrary evidence and its legal implications.
. Nor does the majority convincingly explain the fact that district court made absolutely no reference, in its opinion, to the sophistication of Moore's leadership role in the murder, one of the main elements of the state's argument regarding adaptive functioning. 3 EH 7-11.
. For example, the majority reads the district court as actually deciding that Moore's antisocial tendencies “did not cause his’’ adaptive deficits and that Moore’s role in the crime he committed did not affect its judgment.