joins, concurring in part and dissenting in part.
I concur in the actions of the court today, except that I would reject Bayard Specter’s application for bail pending appeal. Further, I am in agreement with much of what the court says. However, because I believe the majority essentially has rewritten an act of Congress rather than given the act its natural, reasonable interpretation, I express my separate views and, particularly, my understanding of the phrase “likely to result in reversal or an order for a new trial.”
I acknowledge the seductive appeal of the court’s opinion. The logic of its exe*1239gesis, I suppose, may at first glance leave one with every reason not to oppose it. Nonetheless, I believe that to the question the majority poses — “Why not interpret it [i.e., the statutory language] in what may be the most natural and immediately obvious sense, as requiring the defendant to show that the question presented will more likely than not result in reversal?” — the proper answer is that such an interpretation is, in fact, correct. Such a construction not only adheres to the clear language of the statute but also bears out the intent and purpose of Congress.
I have no quarrel with the court’s interpretation of the phrase “substantial question.” To define it as “a close question” or “one that could go either way” is a more workable, practical test than those given in United States v. Miller, 753 F.2d at 23, although the tests there articulated, (“whether the question is “novel,” “not decided by controlling precedent,” or “fairly doubtful”) may have some value in fleshing out the simplicity of the court’s expressed formula.1
However, when the court reaches the second phase of its inquiry, it construes the phrase “likely to result in reversal or * * * a new trial” simply to mean that prejudicial error is required. The court sets out the only legislative history that bears at all on the meaning of this phrase: “The change [in the law] * * * requires an affirmative finding that the chance for reversal is substantial. This gives recognition to the basic principle that a conviction is presumed to be correct.” S.Rep. No. 98-225, 98th Cong., 1st Sess. 27 (1983), U.S.Code Cong. & Admin.News p. 3210. Further, the court correctly observes that the word “likely” “should be read in its ordinary sense as referring to something that is more likely to happen than not.”2 Supra at 1233.
However, the court then simply rewrites the statute in accord with its view of reasonableness by, in essence, inserting the phrase “if decided in defendant’s favor” between the word “likely” and the phrase “to result in reversal or * * * a new trial.” This reduces the meaning of the whole phrase, leaving it to require no more than that release pending appeal should not be granted if the question could be disposed of on a harmless error or a procedural default basis. The approach adopted by the court is simple to apply, relatively certain of result, and possesses the appeal of sweet reason. Its only fault is that it simply is not what Congress has enacted. Further, the approach of the court fails to carry out Congress’s intent that there be “an affirmative finding that the chance for reversal is substantial.”
While the court may be correct in suggesting that to read “likely to result in reversal * * * or a new trial” as I do renders the adjective “substantial” redundant, at least to the extent that questions likely to so result will also be substantial, I do not find this fatal to my interpretation. While it is generally true that interpretations leading to redundancies should be avoided, as the Supreme Court has also *1240stated, the presence of “a clear legislative purpose” may compel such a reading. Singer v. United States, 823 U.S. 338, 344, 65 S.Ct. 282, 285, 89 L.Ed. 285 (1945). Here I believe the legislative history, scant though it is on this point, and “the most natural and immediately obvious sense” of the statutory language establish such a purpose.
I cannot agree with the statement in Miller that such a construction is capricious and would put the district court in the position of a “bookmaker” who “trade[s] on the probability of ultimate outcome.” 753 F.2d at 23. Federal district judges currently make similar determinations in deciding whether a stay of judgment pending appeal should be granted under Fed.R.App.P. 8. We rely on the objective detachment and conscience of a district judge to grant a new trial when he is satisfied that trial error has occurred that will lead to reversal. At other times, however, issues may arise which the district judge may consider not to have such certainty, but which may still suggest the “likelihood of reversal or * * * a new trial.” Congress made clear that the presumption is against release, and it is only the presence of this latter, limited class of issues that may permit the release of a defendant on bail pending appeal.
The court argues that such a position has not been urged by the United States and that we should not adopt “a posture more zealous than that of the prosecution.” If there is zealousness present, it is that of Congress displayed in enacting the statute. The question before us is not that of our posture or that of the prosecution, but what Congress wrote and intended. We should not blithely add a phrase because the attorneys representing the United States urge a meaning of a statute at odds with the statutory language. While deference should be paid to the interpretation of a statute by the agency responsible for administering it, as the Supreme Court has stated, such constructions must be rejected when they “frustrate the policy that Congress sought to implement.” Securities Industry Association v. Board of Governors of the Federal Reserve System, — U.S. -, 104 S.Ct. 2979, 2983, 82 L.Ed.2d 107 (1984) (quoting Federal Election Commission v. Democratic Senatorial Campaign Committee, 454 U.S. 27, 32, 102 S.Ct. 38, 42, 70 L.Ed.2d 23 (1981)).
If one reads the statute straightforwardly, its immediate and apparent sense is simply that following conviction a person may be released on bail if, in addition to the other requirements of the statute, he has raised a substantial question of fact or law that is likely (i.e., more probable than not) to result in reversal or an order for a new trial. What legislative history there is supports this reading. The statute should be applied as written. The second phase of the analysis should consider the issue of the likelihood or probability of error — not just whether harmless, as opposed to prejudicial, error is present.3
Accordingly, I respectfully dissent from that portion of the court’s opinion which, in essence, adds the phrase “if decided in defendant’s favor” to the statute that Congress has enacted. If Congress had felt it desirable, it would have added the language. Congress wished to substantially restrict the number of persons released on bail pending appeal. We should not judicially legislate a meaning more permissive than that clearly apparent from the language of the statute.
. Further, the Miller interpretation is consistent with earlier cases construing the phrase "substantial question” when that was the standard for release pending appeal under the Federal Rules of Criminal Procedure. See, e.g., D’Aquino v. United States, 180 F.2d 271, 272 (9th Cir. 1950).
. In common parlance, if something is "likely” it is "probable” — it has a better chance of happening than not. Webster’s Third New International Dictionary defines "likely” as "of such a nature or so circumstanced as to make something probable” and as "having a better chance of existing or occurring than not: having the character of a probability.” This meaning of "likely” is a common theme that runs through case law as the term appears in often quite different contexts. See, e.g., Munro Drydock, Inc. v. M/V Heron, 467 F.Supp. 513, 515 (D.Mass.1979) (in context of judicial sale of ship, "likely” was synonymous with "probable”); In re Oseing, 296 N.W.2d 797, 801 (Iowa 1980) (for purposes of civil commitment, “likely” means "probable or reasonably to be expected’’); Boland v. Vanderbilt, 140 Conn. 520, 102 A.2d 362, 365 (1953) (in medical testimony, "likely” meant "in all probability; probably”). Both the common understanding of the term and its use in a legal context compel the conclusion that “likely” means "probable” — both terms referring to a greater than 50% chance that something will occur.
. While I do not intend to engage in a full-scale discussion of the constitutionality of the statute, I do observe that the statute addresses a situation where a conviction has been obtained and a defendant’s innocence is no longer presumed. See Stack v. Boyle, 342 U.S. 1, 72 S.Ct. 1, 96 L.Ed. 1 (1951). While blanket denial of post-conviction bail might be improper, I do not believe that the constitution prevents Congress from enacting the restrictions imposed by the clear language of the statute.