dissenting.
The majority’s application of the law-of-the-case doctrine is premised on an invalid assumption: that in considering the “reliability” of the show-up identification in United States v. Singleton, 702 F.2d 1159 (D.C.Cir.1983) (“Singleton I”), this court necessarily passed judgment on the same factors that determine the “reliability,” and hence the admissibility, of the identification for due process purposes. Because a greater degree of reliability is required to determine the admissibility, as opposed to the sufficiency, of identification evidence, I would reach the merits of the admissibility issue.
Due process requires that a suggestive out-of-court identification be excluded from evidence if there is “a very substantial likelihood of irreparable misidentification.” Neil v. Biggers, 409 U.S. 188, 198, 93 S.Ct. 375, 381, 34 L.Ed.2d 401 (1972) (quoting Simmons v. United States, 390 U.S. 377, 384, 88 S.Ct. 967, 971, 19 L.Ed.2d 1247 (1968)). On the other hand, to be sufficient to convict, the identification, in combination with other evidence, need only permit a rational trier of fact to find guilt beyond a reasonable doubt. See Jackson v. Virginia, 443 U.S. 307, 319, 99 S.Ct. 2781, 2789, 61 L.Ed.2d 560 (1979). The latter determination, which was reached in Singleton I, requires a lesser threshold of reliability than the former. In determining sufficiency, the court considers evidence in addition to the identification; accordingly, deficiencies in the reliability of identification can be outweighed by the reliability of other evidence. An admissibility determination, by contrast, requires exclusive focus on the reliability of the identification.
The majority stresses, however, that the identification was virtually the only evidence produced against Singleton. Even in such circumstances, the standard for weighing reliability is stricter when determining admissibility. In weighing the sufficiency of evidence, the judge decides whether the jury could rationally find the identification reliable enough to convict. In weighing the admissibility of evidence, the judge makes his own independent determination of whether the identification is reliable. In the former context, the judge considers the possibility that his own views on reliability are wrong — that a jury could rationally disagree with him and find the identification reliable enough to convict. In the latter context, the judge considers only his own views on reliability: whether *184there is “a very substantial likelihood of irreparable misidentification.” Biggers, 409 U.S. at 198, 93 S.Ct. at 381. Because the required threshold level of reliability need not be discounted by the probability that the judge is wrong, the identification evidence must be more reliable to pass muster in this latter context.
By blurring the distinction between the admissibility and sufficiency of evidence, the majority ignores the most fundamental premises and policies of evidence law. Most rules of evidence, and the due process considerations that often govern those rules, operate to exclude probative evidence. Cf. C. McCormick, McCormick on Evidence § 53 at 135 & n. 2 (E. Cleary 3d ed. 1984). That evidence is probative or even probative enough to convict does not mean it is sound policy to admit the evidence. The policy consideration in the present context is the limits of the rational faculties of factfinders.1 If all jurors conformed to the “rational juror” model of Jackson, there would be no need to exclude as prejudicial a suggestive identification. The rational juror would discount, with precision, the probative value of the identification by the suggestiveness of the surrounding circumstances. But jurors are irrational in the sense that they tend to overvalue the weight of a suggestive positive identification, which is why courts can define such evidence as “prejudicial.” Accordingly, it is the duty of the judge to decide for himself whether the evidence is sufficiently reliable to be admissible. Once the evidence is admitted, the judge merely decides whether the jury could rationally find the evidence sufficiently reliable to convict beyond a reasonable doubt. By interposing the independent judgment of the judge at the admissibility stage, instead of using the deferential Jackson standard at all stages of the proceedings, the Biggers test provides an important safeguard against the jury’s fallibilities. In the long run, unjust convictions will be kept to a sufficiently tolerable minimum to satisfy due process.2
The judgment as to the sufficiency and admissibility of identification evidence, then, is governed by different standards that express different underlying policies. To be sure, in both contexts, the court considers the “reliability” of the identification and weighs the same five indicia of reliability outlined in Biggers. See ante at 181-182. But the extent of the reliability required differs: an identification may be “reliable” enough to convict yet not “reliable” enough to be admissible. In short, the same word has a different meaning in each context. This court’s judgment in Singleton I that the identification, if admissible, was sufficient, along with other evidence, to convict does not preclude a judgment in the case at bar to exclude the same identification evidence on due process grounds.
The majority seems to contend, however, that notwithstanding any theoretical differences between the two reliability determinations, Singleton I in fact decided that the identification was reliable enough for both due process and sufficiency purposes. Yet, the majority in Singleton I expressly declined to reach the due process issue. Singleton I, 702 F.2d at 1166 n. 23. It is true that issues decided as a “necessary implication” of the court's holding are a part of the law of the case. See Bouchet v. National Urban League, Inc., 730 F.2d *185799, 806 (D.C.Cir.1984). A decision as to admissibility may have been necessary in Singleton I in the sense that a defendant can only be convicted on the basis of admissible evidence. But nothing prevented the court from assuming arguendo that the identification evidence was admissible and .then deciding the sufficiency issue; in this way, the district court would be allowed, on remand, to resolve the due process issue in the first instance. This is precisely what the majority in Singleton I decided to do, and it is this determination not to reach the due process issue that is the law of the case.
I recognize that the majority in Singleton I stated that “[njotwithstanding its suggestive elements, the showup at issue here took place under circumstances that would tend to promote reliability.” Singleton I, 702 F.2d at 1166. The necessary implication of this statement is that there was not a “very substantial likelihood of irreparable misidentifieation.” See Biggers, 409 U.S. at 198, 93 S.Ct. at 381. Yet, this statement, and its necessary implication, were not the holding of the case. The court held only that the identification was reliable enough to be sufficient to convict; it expressly refused to reach the due process issue. The extent to which the show-up’s reliability exceeded this threshold showing and any implication that it may have been reliable enough to satisfy due process concerns were dicta. This court has held that dicta is not part of the law of the case. National Souvenir Center, Inc. v. Historic Figures, Inc., 728 F.2d 503, 511 (D.C.Cir.), cert. denied, — U.S.-, 105 S.Ct. 103, 83 L.Ed.2d 48 (1984); accord Gertz v. Robert Welch, Inc., 680 F.2d 527, 533 (7th Cir.1982); Russell v. Commissioner, 678 F.2d 782, 785 (9th Cir.1982); see Quern v. Jordan, 440 U.S. 332, 347 n. 18, 99 S.Ct. 1139, 1148 n. 18, 59 L.Ed.2d 358 (1979) (law-of-the-case doctrine does not preclude consideration of any matter left open by the superior court’s mandate); Vestal, Law of the Case: Single-Suit Preclusion, 1967 Utah L.Rev. 1, 5 (“[ojbviously, the decision of the appellate court has no effect on matters not ruled upon”).
I would therefore decide the due process issue on its merits. For the reasons stated in Judge Wright’s dissent in Singleton I, the identification evidence should be suppressed. Accordingly, I would affirm the order below.
. Often, deterrence is the countervailing consideration that requires the suppression of relevant, probative evidence. The Supreme Court, however, has held that the desirability of deterring suggestive show-ups is not to be considered in determining whether to admit the identification evidence. Manson v. Brathwaite, 432 U.S. 98, 111-14, 97 S.Ct. 2243, 2251-53, 53 L.Ed.2d 140 (1977).
. It could be argued that the rational juror standard of Jackson would adequately protect against this problem. Yet, it is rare that a judge can conclude with certitude that no rational juror could convict on the basis of suggestive identification: the prejudicial effect of such evidence is real, but usually too subtle and too unquantifiable to permit such a conclusion in any particular case. Accordingly, the best safeguard — and, in any event, the safeguard adopted by the Supreme Court — is to rely on the independent judgment of the trial judge that there was not a substantial likelihood of irreparable misidentification.