Bernard F. Bradstreet is the former President and Chief Financial Officer of Kurzweil Applied Intelligence, Inc., a Massachusetts company that develops and sells voice recognition software. Following a twenty-day trial, a jury convicted Bradstreet of conspiring to commit securities fraud, see 18 U.S.C. § 371; securities fraud, see 15 U.S.C. §§ 78j(b), 78ff(a), and 17 C.F.R. § 240.10b-5 ("Rule 10b-5"); and knowingly falsifying Kurzweil's books and records in an attempt to conceal his fraud, see 15 U.S.C. §§ 78m(b)(5), 78ff(a), and 17 C.F.R. § 240.13b2-2. Thereafter, the district court departed downward from the applicable guidelines sentencing range of 51-63 months and sentenced Bradstreet to 33 months in prison, followed by 24 months of supervised release. It also ordered him to pay $2.3 million in restitution.

Bradstreet appeals from his convictions on a number of grounds, only two of which are preserved for plenary appellate review. The government cross-appeals from the district court's sentence, arguing that, on the facts of this case, the downward departure was not within the court's discretion. We affirm the convictions but vacate the judgment and remand for resentencing.

We limit ourselves here to a general overview of the case, deferring more detailed recitations of the facts to later discussions of relevant issues.

To sell stock to the general public on the publicly-traded securities markets, a company must apply for and receive the approval of the Securities and Exchange Commission (SEC), and thereafter make an initial public offering (IPO). In connection with the IPO, the company must file with the SEC a prospectus detailing its overall financial condition and recent financial performance. Subsequently, it also must make quarterly filings of SEC Forms 10-Q, which contain information about the company's financial performance during the preceding quarter.

Sometime in the early 1990's, the Kurzweil management hierarchy, led by Bradstreet, initiated a substantial effort to "take the company public." To this end, Bradstreet established quarterly projections for revenues and profits. Bradstreet then pressured Kurzweil's sales force to meet these projections because investment bankers were unlikely to underwrite the contemplated IPO unless Kurzweil could demonstrate profitability for several quarters in a row.

Companies determine quarterly profits or losses on either a cash or an accrual basis. In cash basis accounting, profit or loss constitutes actual dollars received less actual dollars spent. In accrual basis accounting, profit or loss constitutes revenue due, whether received or not, less expense incurred, whether paid or not. Because informed judgment often determines whether and when revenue actually is "due," public companies that use accrual basis accounting must develop revenue recognition policies that both guide the exercise of such judgment and conform to generally accepted accounting principles (GAAP).

Prior to the decision to go public, Kurzweil, an accrual basis accounter, adopted a revenue recognition policy. In June 1992, management circulated to the sales staff a memorandum reminding the staff of Kurzweil's "policies regarding shipment and revenue recognition." Attached to the memorandum was a document dated "7/28/87" and labeled "Kurzweil Applied Intelligence, Inc. Revenue Recognition Policy." In relevant part, it stated that anticipated revenue should not be recognized if "major uncertainties ... surround culmination of the [revenue-generating] transaction" or if "final acceptance by the customer requires an event out of [Kurzweil's] control...."

After an earlier false start, the IPO closed on August 17, 1993. Thereafter, as required, Kurzweil submitted Forms 10-Q for the quarters ending July 31, 1993 and October 31, 1993. The essence of the government's case was that each of these submissions contained fraudulently-inflated revenue figures indicating that Kurzweil was profitable when, in fact, it was operating near or at a loss. In making its case, the government sought to prove that Bradstreet; Thomas E. Campbell, Kurzweil's vice president in charge of sales; and Debra J. Murray, Kurzweil's treasurer and also a vice president, conspired to and actually did "book" as revenue the anticipated proceeds of a number of contingent sales which occurred in time periods covered by the prospectus and the Forms 10-Q. The government also endeavored to show that these same individuals, along with David R. Earl, Kurzweil's vice president in charge of operations, engaged in a scheme to conceal the fraud from the company's auditors and underwriters. The underlying indictment charged Bradstreet and Campbell with conspiracy (Count I); substantive securities fraud in connection with each of the three fraudulent submissions (Counts II--IV respectively); and knowing falsification of company records (Count V). It also charged Earl with knowing falsification in Count V. Murray had previously entered into a cooperation and plea agreement with the government and had waived indictment.

The indictment set forth 14 improperly-booked "sales" (and alluded to a fifteenth) as overt acts in the conspiracy count. The transactions in question, which took place between June 1992 and January 1994, were of two basic types: (1) those in which, near the end of a fiscal-year quarter, a Kurzweil salesperson had forged a prospective customer's signature to a sales quote; and (2) those in which the prospective customer had signed a sales quote, but had conditioned its agreement to purchase Kurzweil equipment on the occurrence of some event not within Kurzweil's control, such as a future commitment from a third-party purchaser. At trial, the government introduced evidence regarding these transactions and several others, the defendants' knowledge of the nature of these transactions, and the defendants' efforts to conceal the nature of these transactions from Kurzweil's auditors and underwriters. These efforts included the creation of side agreements, not shown to the auditors, which memorialized the conditions of unfinalized sales Kurzweil had recorded as revenue; the forging by Kurzweil personnel of responses to audit "confirmation letters" which the auditors had sent to Kurzweil customers to confirm the details of certain recorded sales; the pretextual shipment of Kurzweil products to a storage facility in order to create, on the books, the illusion of shipment to customers; and the giving of false explanations of the high and ever-growing percentage of Kurzweil revenues made up of accounts receivable. The jury acquitted Earl, but convicted Bradstreet and Campbell on all charges.

Bradstreet's appellate brief presents six developed arguments for reversal of his convictions, but hints at a good number more. As usual, we confine our discussion to the issues accompanied by developed argumentation. See United States v. Bongiorno, 106 F.3d 1027, 1034 (1st Cir.1997).

Because four of Bradstreet's arguments, including the primary one, surface for the first time on appeal, we address them together under the plain-error rubric. See Fed.R.Crim.P. 52(b) ("Plain errors or defects affecting substantial rights may be noticed although they were not brought to the attention of the court."). We then address the two arguments Bradstreet has preserved.

Bradstreet asserts that the trial court plainly erred in failing to give the jury two instructions he never requested. The first is that "the government bears the burden of negating a reasonable interpretation of the revenue recognition policy upon which [its] false statement theory depends" [sic]; the second is that the jury must "unanimously agree on either the factual basis for each count, or the precise legal theory on which [Bradstreet] was guilty" as to the conspiracy and securities fraud counts. Bradstreet further contends that the court plainly erred in permitting the indictment to have been constructively amended and/or in permitting the facts at trial to have varied prejudicially from those alleged in the indictment.

In two recent cases, the Supreme Court has emphasized and then reaffirmed the circumscribed authority Rule 52(b) confers upon appellate courts. To be correctable under Rule 52(b), an error or defect raised for the first time on appeal must be "plain," meaning "clear" or "obvious," United States v. Olano, 507 U.S. 725, 734, 113 S. Ct. 1770, 1777-78, 123 L. Ed. 2d 508 (1993), at the time of appellate consideration, Johnson v. United States, --- U.S. ----, ---- - ----, 117 S. Ct. 1544, 1549-50, 137 L. Ed. 2d 718 (1997); and it must have "affect[ed] substantial rights," meaning, in most cases, "[i]t must have affected the outcome of the district court proceedings," Olano, 507 U.S. at 734, 113 S.Ct. at 1778. Even then, an appellate court should exercise its discretion to notice an error or defect, see id. at 735-36, 113 S.Ct. at 1778-79 (noting the permissive language of the Rule), only if it "seriously affects the fairness, integrity or public reputation of judicial proceedings," id. at 736, 113 S.Ct. at 1779 (citation and internal quotation marks omitted). Although the Court has not described the contours of this discretionary inquiry with much precision, it has declined to exercise its discretion in the face of "overwhelming" evidence that the outcome would have been the same in an error-free proceeding. See Johnson, --- U.S. at ----, 117 S. Ct. at 1550 (involving failure to instruct on an element of the offense). We evaluate Bradstreet's first four appellate arguments against this unfriendly legal backdrop.

1. The Reasonable Interpretation Instruction

The trial court instructed the jury that there are three alternative ways one can commit securities fraud under Rule 10b-5--employing a device, scheme or artifice to defraud; making an untrue statement of a material fact or omitting to state a material fact necessary to prevent the statement made from being misleading; or engaging in an act, practice or course of business which operates or would operate as a fraud or deceit upon any person--and that, although the government need only prove one of these three to secure conviction, the jury's "finding must be unanimous as to which type or types of conduct, if any, have been proven beyond a reasonable doubt." It also told the jury that, to convict Bradstreet of securities fraud, it had to find that he engaged in the fraud knowingly, willfully, and with the intent to defraud. The court then defined for the jury each of these concepts, and concluded its intent instructions as follows:

Because the crimes charged in the indictment involve a specific or deliberate intent to defraud, a good faith belief on the part of a defendant in the truth of his actions or statements will necessarily negate that intent. Even false statements or omissions of a material fact do not constitute a violation of the criminal provisions of the securities fraud law unless made with an intent to defraud. This intent, as I told you, is one that the government must prove beyond a reasonable doubt.

If you were to have a reasonable doubt as to whether a defendant made an inaccurate statement while honestly believing that statement to be true, he cannot be held criminally liable for that statement, even if the statement has been shown demonstrably false. Good faith is a defense to a crime containing an element of specific intent even if a defendant's belief in the proof [sic] of his statements was one that a reasonable person would not have embraced.

Bradstreet did not object to these instructions or seek additional mens rea instructions.

Nevertheless, Bradstreet now contends that the court plainly erred in failing to instruct the jury that Bradstreet would not have committed securities fraud if, in fact, the revenue he knowingly booked was properly booked under any reasonable interpretation of Kurzweil's revenue recognition policy. Analogizing to a false statement case from the Tenth Circuit, see United States v. Migliaccio, 34 F.3d 1517 (10th Cir.1994), and cases cited therein, see id. at 1525, Bradstreet asserts that there is here the possibility that the jury convicted him for one or more recognitions of revenue that were, in fact, reasonable under a fair construction of Kurzweil's policy. Central to Bradstreet's primary argument are subsidiary contentions that the jury was presented with substantively divergent summaries of Kurzweil's policy in the documentary evidence and in the testimony of several witnesses, and that the trial judge never told the jury which version was controlling. Bradstreet also emphasizes our inability to ascertain which transactions the jury relied upon in reaching its verdicts, and our putative willingness to look "more tolerantly" on a "failure to articulate precisely the shape of [a] necessary protective instruction" in the context of an unprecedented prosecution. See United States v. Sawyer, 85 F.3d 713, 742 (1st Cir.1996) (involving a bribery prosecution under the federal Travel Act, 18 U.S.C. § 1952).

The government responds by denying the premises of Bradstreet's argument. It contends that Bradstreet presented as his defense theory lack of knowledge of the fraud, not truth-in-conduct; that all witnesses' summaries of Kurzweil's revenue recognition policy were essentially consonant; that this prosecution was not nearly so novel as the one reviewed in Sawyer; and that Bradstreet's utter failure to argue for a reasonable interpretation instruction below is not comparable to the more forgivable "imprecise articulation" of the argument at issue in Sawyer.

While we agree with the government's final two rejoinders, we think the first two are seriously misleading. On our reading of the record, Bradstreet presented a bifurcated defense. As to the vast majority of the transactions at issue, he denied knowing the critical incriminating facts. But certainly with respect to three transactions--contemplated sales to Transquick, Chicago Mercy, and Willard Hall--and probably with respect to two others--contemplated sales to HCA Nashville and HCA Plano--he did defend on the basis that the revenue that these transactions would have generated was properly recognized under Kurzweil's policy.

Moreover, the jury did hear verbal descriptions of Kurzweil's revenue recognition policy which, when taken in isolation, appear to have differed materially from the written versions of the policy set forth in the trial exhibits. As we have noted, the written version of the policy that was circulated internally at Kurzweil stated that revenue should not be recognized if "major uncertainties ... surround culmination of the [revenue-generating] transaction" or if "final acceptance by the customer requires an event out of [Kurzweil's] control...." See supra at 49. The jury also had before it notes to financial statements, which had been attached to the prospectus, that contained a summary of Kurzweil's policy. In pertinent part, these notes stated: "Revenue from product sales is recorded at the time of shipment if no significant obligation relating to the sale remains and collection is deemed probable." Arguably, these documentary summaries are consistent with one another, and with the synopsis of the revenue recognition inquiry Bradstreet himself presented to the jury: "Are there any major uncertainties and is collection probable?"

They are not, however, entirely consistent with the explanations of the applicable revenue recognition principles provided by two of the government's more important witnesses: Debra Murray, Kurzweil's treasurer, and Harvey Creem, who led Kurzweil's auditing team up to and through the IPO. Both Murray and Creem used language which might suggest that the applicable principles were stricter than the written versions of the policy seemed to indicate.

After being shown a copy of Kurzweil's internal policy, Murray described it as requiring that a "firm contract [exist] before any goods could be shipped"; that the goods "be shipped to the customer and stored at a warehouse only at the request of a customer and that they were going to be paying for the storage [sic]"; and that "there ... be no obligations beyond the company's control." She also noted that Kurzweil's policy was in compliance with GAAP. Creem framed his testimony in terms of GAAP, and not Kurzweil's written policy, stating that income must be "earned" and "realizable" to be recognizable: "Putting that into Kurzweil's terms, Kurzweil would have delivered to a customer a product that the customer wanted, and the customer has the ability to pay and is obligated to pay, both."

If Bradstreet had argued that there was an interpretation of Kurzweil's revenue recognition policy that differed materially from the government's and under which certain of the recognitions of revenue at issue in this case would have been proper, the trial court, upon request, might well have been obliged to give some sort of "reasonable interpretation" instruction. After all, where the government must prove, as an element of the offense, falsity or, as here (at least with respect to the second and third of the three securities fraud scenarios described by Rule 10b-5, see supra at 50-51), something akin to falsity; where the government also must prove intent to defraud; where a defendant advances an understanding of the principles by which truth and falsity are judged that differs from that of the government; and where the defendant's actions might have been truthful under such an understanding, the government cannot carry its burden without first demonstrating the unreasonableness of the contrary understanding. See Migliaccio, 34 F.3d at 1522-25.

In this case, however, Bradstreet never affirmatively claimed, either in testimony or in argument, that his underlying understanding of Kurzweil's policy differed from that of Harvey Creem, Debra Murray, or the government. Nor did he suggest that ambiguities in the policy made such a contrary understanding possible. He merely testified that, in his judgment and on his view of the hotly-contested facts, certain of the transactions put in issue by the government properly triggered a recognition of revenue under Kurzweil's policy. In light of this, we take Bradstreet's characterization of Kurzweil's written policy, a document to which his lawyer drew his attention just before he gave his characterization, to be only a synopsis of the document. We do not take it to be a de facto assertion that Bradstreet's baseline understanding of the policy differed from that of the government, or that a contrary understanding was possible. And absent such an assertion, there was no need for the instruction Bradstreet now contends was necessary.

The transcript demonstrates that the parties tried this case on disputed historical facts and the inferences to be drawn from those facts. The principles underlying the policy by which Bradstreet's conduct was to be judged, though summarized variously and, perhaps, carelessly, were not controverted; they seem to have been commonly understood. This is enough to differentiate this case from Migliaccio and the cases on which it relies. And it is enough to convince us that the trial court's failure to give a sua sponte reasonable interpretation instruction was not plain error. On this record, there is no basis for concluding that the jury's verdict would have been different had the trial judge given the now-suggested instruction. Cf. Johnson, --- U.S. at ----, 117 S. Ct. at 1550.

Bradstreet's remaining claims of plain error merit less discussion. As we have observed, the trial judge informed the jury that it must unanimously agree upon which of the three types of securities fraud Bradstreet committed. See supra at 50-51. In view of this, we are at a loss to comprehend Bradstreet's suggestion that the jury never was told to agree on a precise legal theory of guilt as to the securities fraud and conspiracy counts.

With respect to the argument that the jury should have been told that it must "unanimously agree on ... the factual basis for each count," we simply note that it is unaccompanied by citation to any case which even remotely supports it, and that, although this area of the law is still developing, the weight of the relevant authority appears to be against requiring juries to reach factual unanimity in circumstances such as these. See McKoy v. North Carolina, 494 U.S. 433, 449, 110 S. Ct. 1227, 1236-37, 108 L. Ed. 2d 369 (1990) ("Plainly there is no general requirement that the jury reach agreement on the preliminary factual issues which underlie the verdict.") (Blackmun, J., concurring) (footnote omitted); United States v. Tipton, 90 F.3d 861, 885 (4th Cir.1996) (unanimity instructions need guard only against a lack of unanimity as to the means by which a statute was in fact violated), cert. denied, --- U.S. ----, 117 S. Ct. 2414, 138 L. Ed. 2d 179 (1997); United States v. Bellrichard, 62 F.3d 1046, 1049 (8th Cir.1995) (similar), cert. denied, 517 U.S. 1137, 116 S. Ct. 1425, 134 L. Ed. 2d 549 (1996); United States v. Tarvers, 833 F.2d 1068, 1074 (1st Cir.1987) (unanimity generally not required with respect to a specific act underlying an element of a charged offense); cf. United States v. Shaoul, 41 F.3d 811, 818 n. 4 (2d Cir.1994) (quoting pattern unanimity instructions). We therefore discern no "clear" or "obvious" defect in the trial court's unanimity instructions. See Olano, 507 U.S. at 734, 113 S.Ct. at 1777-78.

We are left, then, with Bradstreet's claims of constructive amendment and/or prejudicial variance. See United States v. Fisher, 3 F.3d 456, 462-63 (1st Cir.1993) ("A constructive amendment occurs when the charging terms of the indictment are altered, either literally or in effect, by the prosecution or court after the grand jury has last passed upon them. A variance occurs when the charging terms remain unchanged but when the facts proved at trial are different from those alleged in the indictment.") (citations and internal quotation marks omitted). Bradstreet first complains about the government's introduction into evidence of transactions other than those set forth as overt acts in Count I of the indictment. He also contests the court's instruction to the jury that it could convict Bradstreet on Count V under an aiding and abetting theory. Finally, he points to the discrepancy between the relatively strict summary of Kurzweil's revenue recognition policy set forth in paragraph 1(f) of the indictment--"[Kurzweil] could only recognize a sale as revenue for purposes of its financial statements and balance sheet when (1) it had a firm, unconditional contract with the buyer evidenced by a signed purchase order or sales quote signed by the customer and (2) it had shipped the product to the customer"--and the more open-ended language found in the written versions of the policy the jury saw. None of these alleged defects is within the purview of Rule 52(b).

First, it is settled that the government "need not recite all of its evidence in the indictment, nor is it limited at trial to the overt acts listed in the indictment." Fisher, 3 F.3d at 462 n. 16 (citation and internal quotation marks omitted). Bradstreet has not pointed us toward anything that takes this case outside the general rule. Second, Count V did effectively charge him with falsifying books and records and aiding and abetting such a falsification by alleging a violation of the aiding and abetting statute, 18 U.S.C. § 2. And even had it not done so, "aiding and abetting is an alternative charge in every count, whether explicit or implicit." United States v. Oreto, 37 F.3d 739, 751 (1st Cir.1994) (citation and internal quotation marks omitted), cert. denied, 513 U.S. 1177, 115 S. Ct. 1161, 130 L. Ed. 2d 1116 (1995). Third, while the summary of Kurzweil's revenue recognition policy set forth in the indictment--a summary the jury did not hear or read--did differ materially from the language used in the written versions of the policy that were in evidence, the lack of congruence did not constructively amend the indictment and cause Bradstreet to be convicted of a crime other than the ones charged. Cf., e.g., United States v. Fletcher, 121 F.3d 187, 191-93 (5th Cir.) (analyzing the effects of a constructive amendment), cert. denied, --- U.S. ----, 118 S. Ct. 640, 139 L. Ed. 2d 618 (1997). Nor did it prejudice him. See Fisher, 3 F.3d at 463 (an objected-to variance constitutes reversible error only if it results in prejudice). Indeed, the variance we detect worked only to Bradstreet's advantage, as the versions of the policy the jury saw were, if anything, more defense-friendly than the summary of applicable principles set forth in the indictment. Cf. id. at 463 n. 19.

Bradstreet contends that the trial court committed reversible error when, in admitting into evidence the cooperation agreement between the government and Debra Murray, it failed to redact from the document the $10 million loss to investors Murray admitted to having caused. He also argues that the court committed reversible error when, in giving the jury an accomplice witness instruction, it inadvertently failed, despite its having told Bradstreet it would do so at the charging conference, to tell the jury to consider what benefits Murray "hopes to receive" in addition to the benefits she had been promised or had received. The first argument is unconvincing and the second is frivolous.

Near the end of the trial's sixth day, Bradstreet and Campbell argued to the district court that the amount of loss Murray admitted to having caused should be redacted because it was irrelevant, see Fed.R.Evid. 401 and 402, or, even if relevant, was highly inflammatory and therefore excludable under Fed.R.Evid. 403 ("Although relevant, evidence may be excluded if its probative value is substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading the jury...."). The district court rejected this argument, reasoning that the amount of loss was relevant to the materiality of the falsely-recorded revenue figures. At Bradstreet and Campbell's request, the court then instructed the jury that the loss stipulation was between the government and Murray only, and that it should not be viewed as binding upon Bradstreet, Campbell, or Earl.

Although Bradstreet's appellate argument is not entirely clear on this point, we infer that he continues to view the amount of loss as either irrelevant or, if relevant, excludable under Rule 403. The government points out that Bradstreet has not presented a coherent challenge to the district court's reasoning in admitting the evidence, and contends further that the evidence was relevant to Murray's knowledge of the scope of the conspiracy. Alternatively, the government asserts that any error in admitting the loss figure was harmless because "it is highly probable that the error did not contribute to the verdict." United States v. Rose, 104 F.3d 1408, 1414 (1st Cir.), cert. denied, --- U.S. ----, 117 S. Ct. 2424, 138 L. Ed. 2d 187 (1997).

On the one hand, Bradstreet says little in his brief about whether the amount of loss was relevant. On the other, we have some trouble seeing how the amount of loss was relevant to the materiality of the alleged false statements or Murray's knowledge of the scope of the conspiracy. We therefore turn our focus to the harmless-error analysis by assuming error arguendo and asking whether the error was likely to have affected the verdict. We see no such likelihood. The jury was well aware that the IPO netted Kurzweil approximately $24 million. Moreover, the jury was told that a private placement of Kurzweil stock would have netted anywhere from $10-15 million less than an IPO. Finally, a single investor, Scudder, Stevens, and Clark, testified without objection that, by April 1994, it had invested approximately $5.6 million in the company. The jury therefore could hardly have been shocked by evidence that the total loss was $10 million. We are confident that this evidence had no effect on the verdict.

As to Bradstreet's objection to the accomplice witness instruction, we think that, although the court failed to use the "hopes to receive" language Bradstreet requested, the court's lengthy instruction was adequate to convey to the jury the need to scrutinize Debra Murray's testimony with special care. This, in combination with the extensive cross examination of Murray as to the benefits she hoped to receive for her plea and cooperation, leaves us with no doubt whatsoever that the jury fully understood it was to regard what Murray had to say with some skepticism. Cf. United States v. Newton, 891 F.2d 944, 950 (1st Cir.1989) (rejecting a challenge to a court's failure to give an accomplice witness instruction because the court's immunized-witness instruction advised the jury to receive the testimony of such a witness with caution and to weigh it with care).

Having rejected Bradstreet's challenges to his convictions, we turn now to the government's cross-appeal. Appropriately applying the 1995 Guidelines Manual, the probation officer who prepared Bradstreet's presentence report (PSR) recommended a base offense level of six; a two-level increase for more than minimal planning; a fifteen-level increase because the loss ($11,471,250.00) exceeded $10 million; a four-level increase because Bradstreet was an organizer or leader of a criminal activity that involved five or more participants or was otherwise extensive; and a two-level increase for abuse of a position of public or private trust. This yielded adjusted and total offense levels of 29 and, because Bradstreet had no criminal history, a recommended guidelines sentencing range of 87-108 months.

Prior to sentencing, however, Bradstreet and the government entered into a sentencing agreement which mirrored the PSR except in two respects. First, the government agreed not to seek a two-level upward adjustment for abuse of a position of trust. Second, the parties agreed to request the court to find that the $11-plus million loss figure overstated the seriousness of the offense, see Application Note 7(b) of U.S.S.G. § 2F1.1, and that the appropriate amount of loss to be attributed to Bradstreet was approximately $2.3 million. Adoption of this calculation would result in a twelve, rather than fifteen, level increase for amount of loss. These provisions of the agreement combined to reduce the recommended total offense level to 24 and the recommended guidelines sentencing range to 51-63 months. The sentencing agreement also provided that the only ground on which Bradstreet could move for a downward departure was under a theory that his conduct was "a single act of aberrant behavior," see United States v. Grandmaison, 77 F.3d 555, 560-64 (1st Cir.1996) (explicating the contours of this ground of departure), and that the government would oppose the motion. Prior to sentencing, Bradstreet so moved.

The government opposed Bradstreet's motion on three grounds. First, it argued that it is illogical to find aberrant conduct where, as here, there has been no admission of guilt. Alternatively, it asserted that both the record and the jury's verdicts establish that Bradstreet testified dishonestly when he testified that he did not act with an intent to defraud, see supra at 50-51 (outlining the court's mens rea instructions, which emphasized that the jury must find an intent to defraud in order to convict); see also, e.g., United States v. Rostoff, 53 F.3d 398, 413 (1st Cir.1995) (a court is bound to accept a fact necessarily established by a jury verdict when that fact is material to sentencing), and that it is illogical to find criminal dishonesty aberrant where the defendant subsequently testified dishonestly. Finally, the government took the position that the duration, complexity, and sophistication of Bradstreet's fraud defy characterization as "a single act."

The district court accepted the parties' recommendations as to the appropriate guidelines calculations, finding that Bradstreet had a total offense level of 24 and an applicable guidelines sentencing range of 51-63 months. The court then granted Bradstreet's motion for a downward departure, reduced Bradstreet's total offense level to 20 (yielding a guidelines sentencing range of 33-41 months), and sentenced Bradstreet to 33 months in prison. In doing so, the court implicitly rejected the government's argument that a defendant must admit guilt in order to receive an aberrant conduct departure. The court also rejected without explanation the argument that the record and verdicts establish that Bradstreet testified dishonestly, and that this fact makes him legally ineligible for an aberrant conduct departure.

Rather, the court looked to our statement in Grandmaison that "aberrant behavior departures are available to first offenders whose course of criminal conduct involves more than one criminal act," 77 F.3d at 563, and our directive that courts judge aberrance vel non under a totality-of-circumstances test, see id. at 563-64 (approving consideration of factors such as the absence of pecuniary gain to the defendant, prior good deeds, and efforts to mitigate the effects of the crime), to find that Bradstreet had engaged in "behavior ... animated by a single objective, ... the success of the Kurzweil IPO." In the court's view, Bradstreet's conduct was, under the facts of this case, tantamount to a single act. And the totality of the circumstances--a perceived lack of motivation by greed, an otherwise exemplary life, a record of significant charitable giving, and an impressive outpouring of support from friends and family--warranted the conclusion that Bradstreet's conduct was aberrant.

Even if we were to follow the district court's approach and to define Bradstreet's criminal conduct at an exceedingly high level of generality, that is, as a multi-faceted act of dishonesty designed to obtain for Kurzweil badly-need cash during the 1992-94 time frame, we are faced with the government's arguments that what occurred was not a single aberrant act of dishonesty because Bradstreet did not plead guilty and/or because Bradstreet engaged in the wholly-separate act of testifying dishonestly about his conduct. Because we see no convincing response to the latter of these two arguments on the facts of this case, we accept it and leave to another day consideration of whether an admission of guilt is a prerequisite to an aberrant behavior departure.

Although Grandmaison takes an expansive view of that which constitutes a single act of aberrant conduct, it confirms that the Guidelines Manual means what it says: a departure for an act that is composed of a number of component acts, id. at 563 ("[S]ingle acts of aberrant behavior ... include multiple acts leading up to the commission of a crime."), is permissible only if the act is singular, see id. at 564 (first time offenders who have been "convicted of several unrelated offenses" are not entitled to aberrant conduct departures). Moreover, in the context of guidelines sentencing, we think it obvious that the term "aberrant" must look forward as well as backward. In other words, an aberrant behavior departure is not warranted unless the conduct at issue is both a marked departure from the past and is unlikely to recur. Cf. United States v. Lam, 20 F.3d 999, 1004 (9th Cir.1994) ("[I]n this context, calling a consistent criminal's behavior aberrant would be an oxymoron and, perhaps, make us look like oxen or morons or both."). In so holding, we note that the Ninth Circuit, which also takes an expansive view of that which constitutes a single act of aberrant behavior, see United States v. Takai, 941 F.2d 738, 741 (9th Cir.1991), apparently includes likelihood of recurrence as part of its aberrance calculation, see Lam, 20 F.3d at 1005.

Under these criteria and on this record, the district court exceeded its discretion in rejecting the government's dishonest testimony argument and departing downward. The argument rests on two premises, one legal and one factual: (1) one convicted of criminal dishonesty who testifies dishonestly about his conduct is not entitled to an aberrant conduct departure as a matter of law; and (2) a finding that Bradstreet did not testify dishonestly would be an abuse of discretion. Because the court failed to specify which of these premises it did not accept, we examine each in turn.

We think it obvious that the government's legal premise is sound. As we have observed, a departure based on a finding that the relevant criminal conduct was a single act of aberrant behavior is appropriate only where the conduct was isolated and is unlikely to recur. Yet one who testifies dishonestly after engaging in felonious dishonesty cannot credibly make either claim. One convicted of criminal dishonesty is therefore not entitled to an aberrant conduct departure if he has testified dishonestly about his criminal conduct.

We also agree with the government's factual premise. As the government pointed out both below and on appeal, Bradstreet testified that he did not intend to file false information in connection with the public offering, to file false financial statements in connection with the relevant Forms 10-Q, or to conceal records or information from the auditors. The verdicts against him necessarily establish, however, that the jury rejected this testimony and found that he did act with an intent to defraud. See supra at 50-51 (noting that the court instructed the jury to acquit unless it found that Bradstreet acted with an intent to defraud and setting forth the court's mens rea instructions). In our view, this finding conclusively establishes that Bradstreet testified dishonestly at trial. After all, the jury's verdict must be credited over Bradstreet's contrary testimony, see, e.g., Rostoff, 53 F.3d at 413; the contrary testimony strikes us as inherently not subject to characterization as unintentional, cf. United States v. Dunnigan, 507 U.S. 87, 94, 113 S. Ct. 1111, 1116, 122 L. Ed. 2d 445 (1993) (making clear that false testimony is not perjurious where it is "a result of confusion, mistake, or faulty memory"); and, in any event, Bradstreet has not responded to the government's argument by suggesting that his false intent testimony was unintentional. To the contrary, he has steadfastly maintained that he acted without an intent to defraud during the entire pendency of these proceedings.

Bradstreet attempts to rebut this line of analysis in three ways. First, he appears in some places to argue that the district court departed downward on some ground or grounds other than the guidelines-based single act of aberrant behavior ground, and that the Supreme Court's decision in Koon v. United States, 518 U.S. 81, 116 S. Ct. 2035, 135 L. Ed. 2d 392 (1996) (establishing an across-the-board abuse of discretion reviewing standard for sentencing departures), validates the court's authority to engage in such a departure. We think it apparent, however, that the court based its departure on the ground on which departure was sought: that the conduct underlying the conviction was a single act of aberrant behavior. To the extent that the court ranged far and wide in explaining its departure, we perceive it only to have been employing the totality-of-circumstances test we prescribed in Grandmaison.

Second, Bradstreet seems to contend that Koon precludes appellate courts from establishing the contours of mixed fact/law concepts such as that which constitutes a single act of aberrant behavior. Koon makes clear, however, that the appellate courts are to continue to establish the legal boundaries and to correct law-based misapplications of such concepts. See 518 U.S. at ---- - ----, 116 S. Ct. at 2047-48. Here, for the reasons just stated, we think the sentencing court went beyond its legal boundaries when it concluded that the dishonest conduct underlying Bradstreet's convictions was both a one-time occurrence and an aberration. We simply have corrected the court's error.

Finally, Bradstreet contends that the jury did not necessarily reject any aspect of his testimony. In doing so, he reanimates his argument that, because the jury heard substantively divergent versions of Kurzweil's revenue recognition policy and was not told to acquit if it found that the revenue Bradstreet knowingly booked was properly booked under a reasonable interpretation of the policy, his conviction is fatally flawed. In Bradstreet's view, the jury might have believed that he knew nothing about the true nature of those transactions involving forgeries, but nonetheless convicted him on the basis of those transactions he defended as having generated properly-recognized revenue.

Even if we assume this unlikely scenario for the sake of argument, it remains fact that Bradstreet never argued that there was an interpretation of Kurzweil's policy that differed in some respect from the government's. And on this record, there is no basis for an inference that the jury understood the testifying witnesses' summaries of Kurzweil's revenue recognition policy to be anything other than divergent synopses of commonly-understood concepts. We thus have every confidence that the jury determined that Bradstreet acted with an intent to defraud by reference to a common and proper set of principles. As a result, we are bound to credit the jury's intent finding, which conclusively demonstrates its rejection of Bradstreet's testimony.

We wish to be clear on the precise nature of our ruling. We do not employ a per se rule that an accused who gives testimony that is necessarily rejected by the jury has intentionally testified dishonestly--i.e., that he has perjured himself. As we have stated, such testimony, though it must be taken as false, see Rostoff, 53 F.3d at 413, may not have been intentionally false; it may have been the product of confusion, mistake, or faulty memory, see Dunnigan, 507 U.S. at 95, 113 S.Ct. at 1116-17. Here, though, for reasons we have explained, see supra at 56-57, Bradstreet's false testimony simply is not capable of being regarded as unintentional.

Because the record is fully developed on this point and Bradstreet has had an ample opportunity to respond to the government's argument, we rule, as a matter of law, that the dishonest activity for which Bradstreet stands convicted was not a single act of aberrant conduct. Accordingly, we vacate Bradstreet's sentence and remand for resentencing. See, e.g., Rostoff, 53 F.3d at 413-14.

Our decision to nullify the district court's downward departure might strike some as harsh. We are acutely aware that incarceration is but one of a number of ruinous consequences that the 52-year-old Bradstreet and his family are suffering as a result of his conduct. And we have a great deal of respect for the informed judgment of the experienced judge who determined that, in light of all the circumstances, nearly three years in prison is enough. But it hardly bears repeating that, under guidelines sentencing, a judge has limited discretion to depart from an applicable guidelines sentencing range. This case is yet another striking reminder of this fact.

For the reasons stated, we affirm Bradstreet's convictions but vacate the judgment and remand for resentencing.