Eugene Hannigan appeals from his conviction for one count of mail fraud, in violation of 18 U.S.C. Sec. 1341. Because there was insufficient evidence produced at trial that the United States mails were used to accomplish the alleged fraud, we will reverse the judgment of the district court and direct that a judgment of acquittal be entered.

Hannigan was indicted on two counts of mail fraud. The jury found him guilty of Count One and not guilty of Count Two. Although Hannigan has raised numerous points on appeal, we will address only those facts and issues concerning Count One dealing with the sufficiency of evidence as to mailing.

Hannigan was the manager of an auto body shop, Park Auto Body, located in Philadelphia. Count One charged that Hannigan and David Giordano, an appraiser employed by Travelers Insurance Company ("Travelers"), submitted a fraudulent insurance claim, falsely representing that a car had been damaged by chemical emissions from a refinery, the Sun Oil Company ("Sun Oil"). The indictment charged that Giordano and Hannigan, "knowingly cause[d] to be delivered by the United States Postal Service ... a $4,001.13 check payable to Park Auto Body on the [false] claim, from Travelers to Park Auto Body." App. at 9.

The prosecution attempted to establish the mailing through a single witness, Cindi Skowronski, a Travelers' claims supervisor. Since Skowronski was the only witness who testified as to the mailing, we will describe her testimony in some detail. Skowronski testified that she assisted in the processing of Sun Oil claims for Travelers, and described at trial the procedures which Travelers followed for processing these claims. She testified that after receiving notice of a claim, Travelers set up an appraisal site or sent appraisers to inspect the damage caused by emissions at the Sun Oil plant, and the appraisers brought their estimates to Travelers' office. After Travelers set up a claim number and subfile for each claimant, it paid the claims by check, often payable to body shops or car rental companies rather than individuals.

Skowronski testified that on a daily basis, Travelers issued checks. She stated: "Within our office, there's a person in charge of running the checks so you couldn't input a check or--during [sic] that time. And, then once they were run off of a printer, they would then be stuffed into envelopes and mailed." App. at 169 (emphasis added). On occasion, however, individuals would arrange to pick up a check at the Travelers office, rather than having it mailed to them. In such a situation, Ms. Skowronski testified to a different procedure:

In order for a check to be picked up at our office ... we would have to have our unit manager approve someone coming in to pick up the check for a check to be released to me. And, proof of that--of them approving it, would be signing the file or signing a piece of paper that was attached to the file. And, then once that was done--when you input the check on the computer, there was a little sign--a little question that said, like check attachment and you would put a yes, so that they know to give me that check. If someone came to pick it up, then I would have it already [sic] ready for them.

App. at 170.

In addition to Ms. Skowronski's testimony--that Travelers usually mailed claim checks and that special procedures were required when someone wanted instead to pick up a check--the government introduced computer printouts for the Sun Oil claims. The computer printouts contained a space entitled "attachment," in which a "Y" or "N" would be placed. Skowronski testified that a "Y" meant the check was authorized to be picked up and an "N" meant that the check was to be mailed. The computer printout for the repair claim addressed in Count One contained an "N" in the attachment column, and Skowronski testified that this indicated that the claim check was to be mailed, not picked up.

On cross examination, Hannigan's counsel engaged in the following colloquy with Ms. Skowronski:

Q: Now, you didn't mail the checks in this case yourself, did you?

A: No.

Q: All right. And, you didn't see them put into the mail yourself, did you?

A: No.

Q: And, can you tell the jury where they're put to be mailed or who mails them?

A: No.

Q: You don't know that?

A: I ...

Q: After they're stuffed in an envelope, you don't know where the envelope goes?

A: To our mail department.

Q: Your mail department. And where is your mail department?

A: At that time, it was on, like--I think we were on the seventh floor and that was, like, on the fifth floor.

Q: The fifth floor. So, you never saw them actually put in the mail or picked up in the mail, is that right?

A: No.

Q: And, someone could go to the mail department and pick one up and you would never know it even though there was supposed to be a procedure, is that correct?

A: That's correct.

App. at 179-80. The government did not conduct redirect examination of Ms. Skowronski.

In reviewing the verdict of the jury, we view the evidence in the light most favorable to the government. Glasser v. United States, 315 U.S. 60, 80, 62 S. Ct. 457, 469, 86 L. Ed. 680 (1942). If there is substantial evidence to support the jury's determination, we do not disturb the verdict although on that evidence we might not have made the same decision. Id.

Hannigan contends that his conviction cannot stand because the above evidence presented at trial was insufficient for the jury to conclude that the United States mails were used to accomplish the alleged fraud. The essential elements of an offense under 18 U.S.C. Sec. 1341 are (1) the existence of a scheme to defraud; (2) the participation by the defendant in the particular scheme charged with the specific intent1 to defraud; and (3) the use of the United States mails in furtherance of the fraudulent scheme. E.g., United States v. Burks, 867 F.2d 795, 797 (3d Cir.1989). In this appeal, we address only whether sufficient evidence was presented to prove the third element.

It is well-established that evidence of business practice or office custom supports a finding of the mailing element of Sec. 1341.2 Once evidence concerning office custom of mailing is presented, the prosecution need not affirmatively disprove every conceivable alternative theory as to how the specific correspondence was delivered. E.g., United States v. Matzker, 473 F.2d 408, 411 (8th Cir.1973). While the element of mailing may be proven through such circumstantial evidence, we have held that to convict under Sec. 1341, some reference to the correspondence in question is required. Burks, 867 F.2d at 797 ("Although circumstantial evidence may be used to prove the element of mailing ... under Sec. 1341, reliance upon inferences drawn from evidence of standard business practice without specific reference to the mailing in question is insufficient.").

In this case, ironically, the government met the specific reference requirement of Burks, but failed to establish that as a routine practice, the United States mails were used by Travelers. The evidence in the record provided a reference to the correspondence in question: Skowronski testified that the "N" in the attachment column of the computer printout for the repair claim addressed in Count One meant that the claim check was to be sent to the mail room. This case is therefore distinguishable from Burks, in which "no evidence was presented concerning the [relevant] correspondence specifically," 867 F.2d at 797.3

Had the government presented some competent evidence that as a routine practice the mail room sent claims checks through the United States mail, the verdict would be sustained. However, the direct and cross examination of Skowronski reveals that she had no personal knowledge that the routine practice of Travelers was to use the United States mails. Skowronski only testified that she knew that the envelopes stuffed with the claims checks would go to the mail room. Neither she nor anyone else established what the business practice was once the envelopes went to the mail room.

The government contends that Ms. Skowronski's testimony is adequate because it is not necessary to produce a witness who personally deposited the correspondence with the United States mails, nor is it required to have a person who was actually employed in the mail room testify as to the business custom and practice of using the United States mails. The government is correct that "[m]ailing can be prove[n] by office custom without producing as a witness the person who personally placed the letter in [the United States mails]." United States v. Joyce, 499 F.2d 9, 17 (7th Cir.), cert. denied, 419 U.S. 1031, 95 S. Ct. 512, 42 L. Ed. 2d 306 (1974). The government is also right that the business practice may be established by the testimony of anyone with personal knowledge of the business custom and practice; it is not necessary that someone actually employed in the mail room establish this fact. See, e.g., id. at 15-16; Matzker, 473 F.2d at 411; 2 Jack B. Weinstein, Weinstein's Evidence Sec. 406, at 406-19 (1993) ("Proof of custom may ... be utilized even when the person who engaged in the routine practices is unavailable to testify. In cases of mailing, the absence of a requirement that the mailing clerk himself testify obviously accords with business realities."). Nevertheless, Ms. Skowronski's testimony indicates that she had no personal knowledge concerning the routine mailing practices of the mail room, or how mail once delivered to the mail room was thereafter forwarded to the United States mails. In fact, not a single witness with personal knowledge testified that it was the routine practice of Travelers to use the United States mails.

Thus, the jury could only speculate what normally happened to correspondence brought to the mail room. For example, it is quite possible that the mail room used a personal messenger or private delivery service to deliver the type of correspondence at issue here. In United States v. Hart, 693 F.2d 286 (3d Cir.1982), we found that testimony that correspondence had been "sent" was insufficient to support a mail fraud conviction, for the very reason that the word "sent" encompasses means of delivery other than United States mails. Id. at 289. Ms. Skowronski's testimony that the envelope in question went to the mail room was tantamount to her saying that the envelope had been "sent," because there was no testimony that the mail room as a routine practice used the United States mails for delivery, and how that was accomplished. Our holding in Hart is controlling, and we therefore conclude that the evidence is insufficient to support the verdict of conviction.

The government argues that there is other evidence in the record to support a finding that the United States mails were used. Giordano testified at trial that Hannigan informed him that Hannigan had received the check in question. The government contends that Hannigan's statement that he received the check is evidence that it was sent through the United States mails. This argument is without merit. The government's argument rests on the unstated false premise that if someone receives something, he must have received it through the United States mails. According to Giordano's testimony, Hannigan never stated that he received the check in the mail--Hannigan merely stated that he received it. Since there are numerous ways to receive correspondence other than through the United States mails, Hannigan's statement that he received the check does not support a finding of the element of mailing to convict under 18 U.S.C. Sec. 1341.

In United States v. Dondich, 506 F.2d 1009 (9th Cir.1974), a Sec. 1341 case with a similar factual setting to the matter presently before the court, the government argued that since most business letters in general are sent through United States mails, a trier of fact could infer that the United States mails were used, even if the prosecution presented no evidence of the custom of the particular business. Id. at 1010. The Court of Appeals for the Ninth Circuit disagreed. It held that where no evidence of custom and usage of mailing practices was presented to the trier of fact, a conviction under Sec. 1341 could not stand. Id.

We agree with the reasoning of the Dondich court. Because the government presented no evidence concerning the custom and practice of Travelers in the use of the United States mails, there is a void in the government's proof on the element of mailing.4 Since there was insufficient evidence presented at trial, a retrial is precluded by the Double Jeopardy Clause of the Fifth Amendment. E.g., Burks v. United States, 437 U.S. 1, 10-18, 98 S. Ct. 2141, 2147-51, 57 L. Ed. 2d 1 (1978). Accordingly, we will reverse the judgment of conviction and sentence, and remand this matter to the district court with a direction that a judgment of acquittal be entered.

BECKER, Circuit Judge, concurring in the judgment.

I concur in the judgment for the reason that, having failed to adduce any evidence of a mailing, the government did not meet its burden of proving that a mailing occurred by any standard of proof. I also agree with my brethren that United States v. Burks, 867 F.2d 795 (3d Cir.1989), while binding upon us, was wrongly decided, insofar as it holds that the government must prove that a business used the United States mails 100 percent of the time (or greater than 99 percent of the time) in order to establish the mailing element. However, I have a somewhat different view from my brethren why this is so and what we can do about it. Because of the importance of the issue, I feel compelled to write separately.

Additionally, while the majority properly attempts to discredit Burks, it appears that it has left standing its requirement for a "specific reference" to the letter mailed. Since I believe that Burks' specific reference requirement is intimately tied to its holding on probabilistic evidence, a holding the majority discredits, I also must write separately to express my view that Burks' specific reference requirement should not be read, as ostensibly done by the majority, as requiring specific reference evidence in every case.

I. THE PROBABILISTIC EVIDENCE COMPONENT OF BURKS

Although I agree that Burks' denigration of statistical evidence made bad law, I have come to the conclusion that Burks is not plainly inconsistent with prior or subsequent law in the Supreme Court. Turner v. United States, 396 U.S. 398, 415-17, 90 S. Ct. 642, 652, 24 L. Ed. 2d 610 (1970), it is true, held that although some heroin is produced in this country, the vast majority of heroin is imported and hence even when judged by the beyond-a-reasonable-doubt standard, a jury may "infer that heroin possessed in this country is a smuggled drug." The Burks panel did not mention Turner, but that does not seal Burks' fate, insofar as there exists a reasonable basis upon which to distinguish the two cases. The issue the Supreme Court confronted in Turner was the validity of a permissive statutory presumption that heroin was imported into the country and that the defendant knew so. The Supreme Court has yet to address directly the question whether the beyond-a-reasonable-doubt standard, at least with respect to inferences drawn from statistical evidence, is identical when no statutory presumptions are involved.1 Moreover, the Court in Turner sifted through all the (non-statistical) evidence presented in that case, which tended to show that the particular defendant possessed the requisite knowledge. See Turner, 396 U.S. at 416-18 & n. 30, 90 S. Ct. at 652-53 & n. 30.

In Turner, the Court apparently viewed the presumption simply as a device placing upon the defendant the burden of coming forward with contrary evidence, a device passing constitutional muster if the presumption is supported by adequate evidence. See Turner, 396 U.S. at 405-09 & nn. 6 & 8, 90 S. Ct. at 646-49 & nn. 6 & 8. Turner does not paint broadly with a probabilistic brush, and is not binding authority for the conclusion that, absent a statutory presumption buoyed by legislative factfinding, probabilistic evidence standing alone (in the sense of being the only evidence linking a predicate fact, such as the possession of heroin, to a fact to be proved, such as the importation of that heroin) may satisfy the constitutional requirement of proof beyond a reasonable doubt.

I also believe that Victor v. Nebraska, --- U.S. ----, 114 S. Ct. 1239, 127 L. Ed. 2d 583 (1994), is not unavoidably inconsistent with Burks, and hence hesitate to conclude that Victor provides a vehicle for overruling Burks. Victor dealt with uncertainties based on inferences and conflicting testimony, not with uncertainty stemming from "naked statistical evidence." See infra at 899 n. 5. The former understanding and acceptance of the probabilistic nature of fact-finding has a long history in American law, having been explicitly recognized by courts for over 100 years, see, e.g., Peter Tillers, Intellectual History, Probability, and the Law of Evidence, 91 MICH.L.REV. 1465, 1466 (1993) (reviewing BARBARA J. SHAPIRO, "BEYOND REASONABLE DOUBT"" and "PROBABLE CAUSE"E" (1991)), but the latter is a relatively new phenomenon that most courts (including two decisions by this Court) have to date generally rejected, see, e.g., Burks, 867 F.2d at 797; Guenther v. Armstrong Rubber Co., 406 F.2d 1315, 1318 (3d Cir.1969) (ordering a directed verdict for the defendant tire manufacturer although the plaintiff, who was injured by an allegedly defective tire sold by a department store, introduced evidence that the defendant manufactured 75-80% of tires sold by that store) ("[T]here was no justification for allowing plaintiff's case on that so-called probability hypothesis to go to the jury. The latter's verdict would at best be a guess. It could not be reasonably supported."); People v. Collins, 68 Cal. 2d 319, 66 Cal. Rptr. 497, 438 P.2d 33 (1968) (in banc) (reversing convictions based solely on (flawed) statistical evidence that the probability of a blond white woman with her hair in a pony tail being in a yellow car accompanied by a black man sporting a beard, a description which the defendants matched, is 1 in 12 million); Smith v. Rapid Transit, Inc., 317 Mass. 469, 58 N.E.2d 754 (1945) (holding that evidence showing that only defendant's bus was licensed to operate on a given street and that the accident occurred near the scheduled time for defendant's bus to be there was insufficient proof to get to the jury on the question of whether defendant's bus caused plaintiff to veer into a parked car); People v. Risley, 214 N.Y. 75, 108 N.E. 200, 203 (1915) (rejecting use of (unsubstantiated) statistical evidence demonstrating that defendant's typewriter was used to forge a document and stating that probabilistic testimony could not be used to establish a historical fact).2

Facts drawn from testimony or other evidence, and more so for those drawn as inferences about facts drawn therefrom, suffer from the impossibility of certain proof. Epistemology has long given up the notion that any historical event can be known with certainty, and so to that extent all "facts" are probabilistic. To be sure, some facts and inferences approach what for all practical purposes amounts to certainty (for example, the proposition that, because I am alive, I was conceived by a woman). But most inferences drawn at a trial are not as convincing as the example cited. Inferences are persuasive to the extent that experience and reason reveal a strong "causal" or "correlative" relationship between events, meaning that one of the events will be more likely given the existence of the other than given the non-existence of the other (this is essentially the standard of relevance provided by Federal Rule of Evidence 401). When the two events are perfectly correlated throughout the course of a very large number of observations--that is, when one event always occurs in the presence of the other and never occurs in its absence--we can be fairly confident that there is a causal or connected relationship between the two events. An illustration is gravity, which exerts a force accelerating each object toward all other objects' center of gravity. When such a phenomenon is observed to occur often and regularly enough, we treat it as a law of nature, and thereafter are justified in making inferences based thereon that are virtually "certain" (that is, the probability of the event of an unsupported ball in our atmosphere returning to a supported position is about as close to 1 as can be).

Few inferences have the force of laws of nature, however. Most are in reality "probabilistic." Flight from the police, for example, is generally assumed to reflect consciousness of guilt, which in turn is generally assumed to make it more probable that the defendant is guilty of the crime charged, but since innocent people are also known and believed to flee from the police and since people guilty of offenses other than the one with which the defendant is charged are known and believed to flee from the police, that event alone does not establish guilt beyond a reasonable doubt of the particular crime charged. It is precisely the strength of associations between events in society, associations that are more often the product of culture and social norms than not, that a large, popular body such as our jury is empaneled to assess. Cf. IA WIGMORE ON EVIDENCE Sec. 37.3, at 1027 (Peter Tillers ed. rev. ed. 1983) (using the term "assertoric" act or judgment to refer to the exercise whereby a probability is attributed to a fact).3

Facts derived from testimonial forensic evidence suffer from a similar "deficiency." There is always the possibility that a witness misperceived the event,4 does not completely recall the event, does not precisely communicate what was recalled, or does not attempt to communicate what was recalled truthfully. Although cross-examination corrects for these problems, it cannot eliminate them. One can never be completely certain that an event actually happened the way a witness describes it. Again, we have accepted the jury as the best institution to determine facts through a collective decisionmaking process. Nevertheless, facts themselves are only probabilistic in the Victor sense.

As mentioned briefly above, the context of the statement the majority retrieves from Victor was the meaning of beyond a reasonable doubt when there is conflicting evidence. See 465 U.S. at 617-19, 104 S.Ct. at 1245 ("With regard to moral evidence, there is, for the most part, real evidence on both sides. On both sides, contrary presumptions, contrary testimonies, contrary experiences must be balanced." (internal quotation omitted)). In such cases, it is clear that one can generally not be certain who is telling the "truth," or for that matter if anyone is. But, as I have noted, such probabilities are beyond the scope of our present problem. Here we are concerned with the question whether undisputed proof stated in probabilistic terms suffices, not what the probabilities are that disputed evidence which is attested to with confidence and accepted by the jury as true is true. Evidence, such as that rejected by Burks, shares both weaknesses, yet courts generally reject the former and accept the latter.

But if the Supreme Court is willing to accept facts and inferences as resting solely on probabilities, rather than certainty, what difference is there between accepting "naked statistical evidence?"5 One cannot rely too much on the case law, for what little there is of it has not comprehensively grappled with the subject. But what is lacking in case law is made up for by a surprising flurry of scholarship surrounding this issue. See, e.g., IA WIGMORE ON EVIDENCE Sec. 37.1, at 1011 n. 6 (collecting scholarly works).

Many commentators have provided incisive reasons for distinguishing between "covertly" probabilistic fact-finding and inference-drawing, and the use of "overtly" probabilistic evidence.6 Others have come to the opposite conclusion, usually tagging the difference between probabilistic and non-probabilistic evidence as one of form rather than substance.7 While I am inclined to agree with those advocating the more widespread use of probabilistic evidence, see, e.g., Maj. Op. at 893 n.3 (citing authorities), and would urge this Court to reconsider Burks in banc, the strength of the arguments that have been raised for distinguishing the two types of probabilities force me to conclude that Victor is not necessarily incompatible with Burks and hence that it provides no basis for overruling Burks. I will set forth some of the stronger arguments against naked statistical evidence here, although I iterate that in the end I do not find them persuasive.

Burks does not set up a requirement that testimonial or circumstantial evidence be 100% reliable; had Burks done so, it would fall under the weight of Victor. No evidence is 100% reliable, see supra at 897-898, and such foolproof reliability is not needed. But there is reliability, and then there is reliability. Social science research, for example, has established beyond peradventure that witness identifications, especially when cross-racial and based on brief moments of observation, are quite unreliable. Perceptions, memories, communications, and veracity in general are not perfect. Yet such unreliability is not a ground to set aside a jury verdict; for lack of a superior alternative, we trust the jury, however foolishly, to resolve those types of unreliabilities satisfactorily.8

Another problem with reliance on "naked statistical evidence" is evident. The majority seems to set the probability cut-off at 95 percent. See Maj.Op. at 893 n. 3.9 Eventually, given enough cases, we will be called upon to decide definitively the precise probability at which guilt is established beyond a reasonable doubt.10 Unfortunately, once we set a threshold percentage we will have every defendant showing us statistics on the reliability of each piece of evidence, performing numerical operations on the numbers, and demanding acquittal. We will need judges and lawyers with degrees in probability theory to make sense out of the data.

To avoid these sorts of problems, when engaged in appellate review we entertain a certain fiction: we suppose that evidence the jury was persuaded to be true, is. That is what I think both Victor and Holland v. United States, 348 U.S. 121, 140, 75 S. Ct. 127, 137-38, 99 L. Ed. 150 (1954) were getting at. We discount the likelihood that a witness who claims to be sure of something cannot be, or is not, even if that likelihood is quite high, unless it reaches the level of incredulity. Instead we assign to the jury, a sort of safety-net black box, the responsibility to sort out truth from untruth and to draw reasonable inferences from the truth. When an innocent person is convicted, we assign error to the unavoidable foibles of the jury system, and avoid attributing it to the deliberate workings of the justice system.

Public confidence in the judicial system requires courts to draw the line somewhere. Otherwise, a 95% probability (the majority's ostensible minimum percentage) of a sperm match of DNA alone (i.e., assuming not one other bit of evidence) could be enough to defeat an alibi defense, no matter how strong, of an alleged rapist, and similarly a 95% probability of a fingerprint match could standing alone suffice to convict someone of robbery if the jury rejected a strong alibi defense (and even contrary statistical evidence). Moreover, evidence that 95% of drivers on a certain stretch of highway speed would be enough to convict all who use the highway (that is, any random driver arrested) of speeding. The law has not yet progressed to that point, and I think that this is what Burks recognizes. Burks does not set up a requirement of "absolute certainty," but implements a sensible criterion courts use to maintain public respect for and confidence in judicial institutions.11 I respect that other courts disagree on this point, but I do not see the court of appeals cases the majority has cited in support of its statistical evidence approach as having thoroughly reasoned through the point; they merely summarily state the fact that a demonstrated probability of guilt alone is enough.12 I surmise that they might balk at the analogous examples I have just raised.13

The foregoing discussion expresses my basic disagreement with the majority's conclusion that Burks' refusal to accept "naked statistical evidence" can be contained to its four corners. Were it a question of first impression I would not decide Burks the way it was, and would instead consider proof that 99% of all mailings were received through the United States mails sufficient to sustain the jury verdict; but it is my view that Burks is not inconsistent with Supreme Court precedent and hence binding upon us.

I turn then to an explication of my reasons for differing with the majority's reading of Burks regarding the specific reference requirement.

II. THE SPECIFIC REFERENCE REQUIREMENT OF BURKS

I begin this discussion by respectfully noting my disagreement with the majority's conclusion that the government met the specific reference requirement of Burks in this case. In Burks, the defendant was charged with submitting fraudulent medical bills to an attorney, who subsequently submitted them to insurance companies. This court reversed the conviction because there was insufficient evidence that the letter containing the fraudulent insurance claim had passed through the United States mails, an element of the offense. The government adduced testimony from two witnesses circumstantially demonstrating the custom and practice of mailing and receiving correspondence.14

The majority apparently attempts to limit Burks holding to require the government to prove "some [specific] reference to the correspondence in question," and it concludes that here Skowronski's testimony provided such a "specific reference." Maj.Op. at 893. I am unpersuaded by that assertion because I have a different understanding of what Burks meant by specific reference. The "specific reference" in this case--Skowronski's testimony that she in fact forwarded the letter in question to the "mail room"15--is not materially different from the evidence in Burks that the attorney forwarded the letter in question to his secretary (the attorney's "mail room"), or to the testimony by the insurance company representative (Walters) that the letter was in the company's files (and thus had very likely been received by its "mail room").

I will focus on Walters' testimony that 99% of correspondence the office received came via the United States mails, although something similar could be said of the other witness' testimony. Walters had specifically referenced the correspondence in question, because the letter was found in the company's files. For purposes of sufficiency of the evidence, assuming that there is "specific reference" testimony placing the letter in question at point X, I see absolutely no logical difference between the persuasiveness or sufficiency of testimony that 99% of all mail reaching point X (the company's files) came from the United States mails and testimony that 99% of the mail reaching point X (the company's mail room) was delivered to the United States mails. In either case one has a specific reference but only a "probability" that the mails were used, a probability that Burks found deficient as a matter of law. Thus it seems to me that, had the government established here that 99% of the letters reaching the mail room were delivered through the United States mails, the evidence would still have fallen short of establishing guilt beyond a reasonable doubt according to the reasoning of Burks despite the specific reference testimony adduced here.

What the majority fails to acknowledge is that the only point of the "specific reference" requirement in Burks was to boost the probabilities up from 99 to 100%.16 I have no doubt that, had Walters (not incredibly) testified that the United States mails were always used (that is, used "100% of the time"), the Burks Court would have found the evidence sufficient to prove that a mailing had occurred despite the lack of a specific reference, for then the testimony would have established "more than a probability that the mails had been used." Id. 867 F.2d at 797. In short, I believe the specific reference requirement in Burks is intimately linked to its judgment that " '[p]robability is not enough to convict a party of mail fraud,' " id. (quoting Scott, 730 F.2d at 147); the two concepts come to us as a package deal.

Accordingly, it is my opinion that Burks does not require "specific reference to the mailing in question" in future mail fraud cases where the government proves beyond a reasonable doubt that all correspondence was sent via the United States mails. I hope, however, that, in the near future, a case presenting these issues will go in banc so that Burks can be properly interred. In the meantime, I hope that the government will be more careful with its proof (or with its decision as to when to bring mail fraud charges). If it is, the problem for the most part will go away.