Appellant Yu Kikumura was convicted of several explosives and passport offenses, for which the federal sentencing guidelines prescribed a sentencing range of between 27 and 33 months in prison. Evidence produced at the sentencing hearing indicated that Kikumura had manufactured three lethal home-made firebombs from scratch in preparation for a major terrorist bombing on American soil--conduct far more serious than is evident from the face of the charged offenses. The district court imposed a sentence of 30 years' imprisonment--apparently the largest departure from an applicable guideline range, in absolute or percentage terms, since the sentencing guidelines became effective. Kikumura appeals that sentence.

We must decide a number of difficult issues: (1) the appropriate standard of proof with respect to factfinding bearing on departures when the magnitude of a proposed departure dwarfs the guideline range applicable to the substantive offenses of conviction; (2) the admissibility for sentencing purposes of hearsay statements by a confidential informant; (3) whether the circumstances of Kikumura's offense were adequately taken into consideration by the Sentencing Commission in fashioning the guideline range from which the district court departed; (4) how to determine the reasonableness of departures; and (5) whether the sentence imposed in this case was reasonable under the appropriate methodology.

We conclude that: (1) under the circumstances, a clear and convincing standard of proof is required (and was met); (2) in view of extensive corroboration from independent sources, the statements of the confidential informant were sufficiently reliable to be considered; (3) the circumstances of Kikumura's offense were not adequately contemplated by the Sentencing Commission, and hence an upward departure was permitted; (4) a useful methodology for making an offense-related departure from the guidelines (and the one we apply here) is to determine what offense level a defendant's conduct most closely approximates by analogizing to other guidelines; and (5) under the circumstances, an upward departure to 360 months was unreasonable, but an upward departure to at most 262 months would not be unreasonable. We therefore will vacate the judgment of sentence and remand for resentencing.

Before we consider the sentencing issues, we must address Kikumura's contention that the district court erred in failing to suppress evidence which, according to Kikumura, was obtained in violation of the fourth amendment. We find no fourth amendment violation and therefore reject this argument.

The relevant facts, as found by the district court, are as follows.1 On the morning of April 12, 1988, New Jersey State Trooper Robert Cieplensky entered the Vince Lombardi service area at the northern end of the New Jersey Turnpike. Cieplensky noticed a man, later identified as Kikumura, twice begin to walk from his car to the service area, only to return to his car abruptly upon establishing eye contact with the trooper. As Cieplensky drove past, Kikumura "milled around" his car for a time. He later walked back toward the service area restaurants, but when he saw Cieplensky turn down a different traffic lane and head toward the restaurants, he quickly turned around, returned to his car, and once more began "milling around." When Cieplensky passed him again, however, Kikumura entered his car and began to drive away in what Cieplensky considered to be a reckless fashion--cutting across two parking lanes within five feet of parked vehicles at a rate of between 25 and 30 miles per hour. Cieplensky ordered Kikumura to stop. Kikumura did so, exited the car, and met Cieplensky outside.

Cieplensky observed fresh burn marks on Kikumura's neck, as well as bandages on his neck and hands. Also, Cieplensky noticed a large black bag and its contents in plain view inside the car. The bag contained seven empty "Hercules" brand gunpowder canisters and a pouch of lead shot. When full, the canisters would have contained several times the amount of gunpowder Cieplensky's father, a regular hunter, would use during an entire season. Concluding that Kikumura could not have been using that amount of gunpowder simply for recreational purposes, Cieplensky conducted a pat-down search of Kikumura, but found no weapon.

Cieplensky then noticed, also in plain view inside Kikumura's car, a cardboard box containing three red cylindrical objects with black tape and wires. He asked Kikumura what was in the box, and Kikumura replied: "Souvenirs. Go ahead, check it out." When Cieplensky did so, he discovered the bombs, handcuffed Kikumura, and placed him under arrest inside the patrol car. Cieplensky then searched the rest of Kikumura's car, including the trunk.

Kikumura, however, insists that he had carefully hidden his bombs, gunpowder canisters, and lead shot inside a suitcase in the trunk of his car. In other words, Kikumura challenges as clearly erroneous the district court's finding that these items were discovered in the passenger compartment. Kikumura faces an uphill battle on this score, for he must overcome a standard of review under which reversal is permitted only if we are "left with the definite and firm conviction that a mistake has been committed." United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S. Ct. 525, 542, 92 L. Ed. 746 (1948). If the district court's finding is "plausible in light of the record viewed in its entirety," we must accept it, even if we would have evaluated the evidence differently in the first instance. See Anderson v. City of Bessemer City, 470 U.S. 564, 573-74, 105 S. Ct. 1504, 1511-12, 84 L. Ed. 2d 518 (1985).

At the suppression hearing, Kikumura told a story entirely different from Cieplensky's. According to Kikumura, Cieplensky, after finishing the pat-down search outside the car, removed the keys from the ignition without Kikumura's permission and proceeded to open the trunk of the car and the suitcase inside the trunk, only then finding the bag containing the gunpowder canisters and lead shot and the box containing the bombs. Deciding which of these irreconcilable stories to believe involved resolving a swearing contest between live witnesses, which the district court was far better situated to do than are we. We must defer, for example, to the district court's evaluation of the "demeanor, tone of voice, [and] manner of response" of the witnesses, 698 F. Supp. at 552, each of which is essentially unreviewable.

Kikumura, however, relies heavily on the testimony of John Crowley, a gas station attendant who observed part of the encounter between Cieplensky and Kikumura. Crowley testified that he saw Cieplensky looking into the trunk of Kikumura's car while Kikumura stood outside beside the car. Crowley did not observe where Kikumura's bombs were located. However, as Kikumura points out, Crowley's testimony is inconsistent with Cieplensky's to the extent Cieplensky testified that he had placed Kikumura inside the patrol car before searching the trunk.2

Kikumura contends that it was clear error for the district court not to credit the testimony of Crowley, "the one eyewitness with absolutely no motivation to lie." Reply Br. at 21. We disagree. A witness's testimonial capacity depends not only on his truthfulness, but also on his perception and memory: in order to give an accurate account of some past event to a factfinder, the witness must accurately (1) perceive the event when it occurred, (2) remember what he perceived between the time of the event and the time of his testimony, and (3) relate what he remembers at the time of his testimony. Although Crowley surely had the least incentive to lie, he just as surely had the least opportunity to perceive the relevant events and the least reason to remember them accurately.

It is probable that the encounter between Cieplensky and Kikumura failed to make any deep impression on Crowley when he observed it. Crowley testified that he had previously observed "quite a few" cars pulled over and searched by troopers at the service station. When Crowley observed Cieplensky and Kikumura, he was driving at an unusually high rate of speed, hurrying to get to work because he was late. He only managed to "glance" at them "for about two or three seconds." Later in the day, however, Crowley observed a bevy of troopers and bomb specialists ransacking the passenger compartment and open trunk of Kikumura's car, which "made quite an impression" on him. Also, he later observed several pictures of the open trunk in the newspapers and on television. Although there is every reason to believe that Crowley testified truthfully, under these circumstances it is quite possible that Crowley simply misremembered seeing the trunk open when he briefly glanced at Kikumura while driving hurriedly to work. The district court's failure to credit Crowley's testimony on this point is not enough to render its subsequent findings clearly erroneous.

As between Cieplensky and Kikumura, the district court was amply justified in concluding that it was Kikumura who was lying. Kikumura himself admitted to lying repeatedly while entering the country illegally and preparing for his bombing. Moreover, Kikumura had a great incentive to lie--his presumed desire to avoid a long prison term--and he had testified inconsistently about some of the relatively minor details of his story, see, e.g., 698 F. Supp. at 548 n. 4. Also, Kikumura's testimony itself contains no small degree of implausibility--for example, that Cieplensky, following a routine traffic stop and unremarkable pat-down frisk, without permission removed the keys from the ignition, opened the trunk, opened the suitcase, opened the box, and only then found evidence of serious wrongdoing.

As Kikumura demonstrates, countervailing arguments can be marshalled. For example, Cieplensky, as well as Kikumura, might have had an incentive to lie, assuming that the searches were illegal, and one might not expect a trained terrorist to leave his bombs where they could be discovered so easily. Although not frivolous, these contentions hardly leave us "with the definite and firm conviction that a mistake has been committed." United States Gypsum Co., 333 U.S. at 395, 68 S.Ct. at 542. The district court's findings are simply not clearly erroneous.3

Kikumura next contends that Cieplensky committed several fourth amendment violations even assuming the correctness of the district court's factfinding.

First, he asserts that the original stop was unjustified because interpreting New Jersey's careless driving statute so as to encompass Kikumura's driving would render the statute void for vagueness. We disagree. The statute prohibits "driv[ing] a vehicle on a highway carelessly, or without due caution and circumspection, in a manner so as to endanger, or be likely to endanger, a person or property." N.J.Stat.Ann. Sec. 39:4-97 (West 1973). The district court found that Kikumura "cut[ ] across two parking lanes at twenty five to thirty miles per hour within five feet of parked vehicles." 698 F. Supp. at 558. Whatever the ambiguity at the margins of the statute, Kikumura's argument can succeed only if the statute is vague as applied to his specific conduct, see, e.g., New York v. Ferber, 458 U.S. 747, 767-69, 102 S. Ct. 3348, 3359-61, 73 L. Ed. 2d 1113 (1982), and we have no doubt that the statute encompasses his conduct.4 Because Cieplensky, who observed Kikumura's driving, had reasonable suspicion that Kikumura was violating the law, the initial stop was justified. See, e.g., United States v. Sokolow, 490 U.S. 1, 109 S. Ct. 1581, 1585, 104 L. Ed. 2d 1 (1989).

Second, Kikumura contends that Cieplensky was unjustified in conducting the pat-down search for weapons. We disagree. A police officer may search a detained individual for weapons if he has reasonable suspicion that the individual could be armed and dangerous to the officer or others. See Terry v. Ohio, 392 U.S. 1, 27, 88 S. Ct. 1868, 1883, 20 L. Ed. 2d 889 (1968). The reasonable suspicion standard of justification, developed in Terry as a more lenient exception to the usual probable cause standard, requires that the officer, at the time of the search, must know of "specific and articulable facts which, taken together with rational inferences from those facts, reasonably warrant th[e] intrusion." Id. at 21, 88 S.Ct. at 1880. The "specific and articulable facts" asserted to justify the pat-down search in this case include Kikumura's possession of lead shot and containers that once held an extremely large quantity of gunpowder, coupled with Kikumura's fresh burn marks and with conduct at least suggesting a fear of detection.

We believe that the chain of inference from the known facts to the perceived danger here is no more attenuated than in other cases in which the Supreme Court has upheld protective searches under the reasonable suspicion standard. For example, officers can infer that suspects might be armed and dangerous from conduct suggesting that they are contemplating a daytime robbery, see Terry, 392 U.S. at 28, 88 S.Ct. at 1883, and can infer that a car containing a large hunting knife might also contain other dangerous weapons accessible to its occupants, see Michigan v. Long, 463 U.S. 1032, 1051, 103 S. Ct. 3469, 3481, 77 L. Ed. 2d 1201 (1983). In such cases, although the specific facts observed do not in any rigorous sense constitute proof that the officer or others are in danger, the standard necessary to establish reasonable suspicion "is considerably less than proof ... by a preponderance of the evidence." Sokolow, 109 S. Ct. at 1585. Because we believe that Cieplensky had reasonable suspicion to fear that Kikumura could be armed and dangerous, we conclude that the pat-down search for weapons was constitutional.

Finally, Kikumura contends that the search of the passenger compartment of his car, during which Cieplensky discovered the bombs, was unconstitutional. A search undertaken pursuant to voluntary consent is not unconstitutional, however, and the district court concluded that Kikumura's consent was voluntary. That determination is a finding of fact, see Schneckloth v. Bustamonte, 412 U.S. 218, 227, 93 S. Ct. 2041, 2047, 36 L. Ed. 2d 854 (1973), and, as such, is subject only to clearly erroneous review. See, e.g., United States v. Kelly, 708 F.2d 121, 126 (3d Cir.), cert. denied, 464 U.S. 916, 104 S. Ct. 279, 78 L. Ed. 2d 258 (1983); see also id. at 127 (Gibbons, J., dissenting). We cannot conclude that the finding of voluntariness was clearly erroneous.

In Schneckloth, the Court instructed that the voluntariness of a consent to search must be determined by reference to "the totality of all the surrounding circumstances--both the characteristics of the accused and the details of the interrogation." 412 U.S. at 226, 93 S.Ct. at 2047. In reviewing the circumstances here, we are struck by one fact of almost overwhelming significance: Kikumura did not merely assent to a search that was first requested by Cieplensky; he suggested it. When Cieplensky asked Kikumura what the box in his car contained, Kikumura responded, "Souvenirs. Go ahead, check it out."

None of the surrounding circumstances vitiate the obvious inference that Kikumura, hoping that Cieplensky would not call his bluff, volunteered to have his car searched. To be sure, Kikumura had been stopped and frisked. However, this stop and frisk, like the stop described in Berkemer v. McCarty, 468 U.S. 420, 104 S. Ct. 3138, 82 L. Ed. 2d 317 (1984), was effected "by only one ... policem[a]n," occurred outdoors in "public view" and in broad daylight. It thus cannot plausibly be characterized as a " 'police dominated' " encounter, id. at 438-39, 104 S.Ct. at 3149, so as to trigger significantly stricter scrutiny of the consent. Kikumura also notes that he was never informed of his right to refuse consent. That fact is relevant in considering the totality of the circumstances, but is hardly dispositive, as Schneckloth 's rejection of the Ninth Circuit's attempt to Mirandize fourth amendment consents makes clear. See 412 U.S. at 227, 93 S.Ct. at 2047. Finally, Kikumura argues that the consent was tainted by Cieplensky's earlier unlawful conduct. Having found neither the stop nor the weapons frisk unconstitutional, however, we may reject this argument summarily.

Because we cannot say that a totality of the circumstances indicates that the district court's finding of a voluntary consent to search was clearly erroneous, we hold that Cieplensky's search of the passenger compartment of Kikumura's car was constitutional.5

Kikumura was charged with twelve passport and weapons offenses, by far the most serious of which was violation of 18 U.S.C. Sec. 844(d), which prohibits transporting any explosive in interstate commerce "with the knowledge or intent that it will be used to kill, injure, or intimidate any individual or unlawfully to damage or destroy any building, vehicle, or real or personal property."6 After the district court denied Kikumura's motion to suppress the evidence discovered in his car, Kikumura stipulated to all essential facts of all the charged offenses and waived his right to a jury trial. However, Kikumura denied that he intended to use his bombs in order "to kill, injure, or intimidate any individual," 18 U.S.C. Sec. 844(d) (emphasis added), conceding only his intent "unlawfully to damage or destroy any building, vehicle, or real or personal property," id. (emphasis added). Following a pro forma bench trial on the stipulated facts, Kikumura was convicted of all counts.

Because Kikumura's offenses were committed after November 1, 1987, his sentencing was governed by the Sentencing Reform Act of 1984, Pub.L. No. 98-473, 98 Stat.1987 (1984), and the sentencing guidelines promulgated thereunder. The parties do not dispute the application of the guidelines outlined in the presentence report. The offense level for Kikumura's violation of 18 U.S.C. Sec. 844(d) was calculated under Sec. 2K1.6 of the guidelines and determined to be 18.7 Because Kikumura's other weapons offenses involved substantially the same harm as his Sec. 844(d) offense, and because his passport and visa offenses were all significantly less serious than his Sec. 844(d) offense, Kikumura's combined offense level was also determined to be 18.8 With no prior convictions, Kikumura was classified as having a criminal history category I, which resulted in a sentencing range of between 27 and 33 months.9 The government's sentencing memorandum put Kikumura on notice that it would seek a substantial upward departure based on Kikumura's alleged terroristic activities.

The district court conducted a sentencing hearing on February 7, 1989. Despite the opportunity to do so, Kikumura declined to testify, to call live witnesses, or to introduce affidavits or other evidence on his behalf. He did, however, vigorously cross-examine both witnesses who testified for the government, and he objected to the introduction of an affidavit relied upon by the government to link Kikumura to the Japanese Red Army (JRA), a notorious international terrorist organization.

Detective Joseph Fuentes, a New Jersey police officer assigned to investigate Kikumura after his arrest, chronicled Kikumura's escapades in entering the United States and preparing for his bombing. On appeal, Kikumura does not seriously contest the details that testimony.

On February 29, 1988, Kikumura visited the American embassy in Paris and presented a Japanese passport in the name of one Masatoshi Kishizono. The passport had originally been issued to Kishizono, was stolen from him, and was skillfully altered so as to replace Kishizono's picture with Kikumura's. Using Kishizono's passport, Kikumura obtained an entry visa to the United States.

On March 8, 1988, Kikumura entered the United States. Soon thereafter, he rented a Manhattan apartment, purchased a used car, and began a four-week odyssey during which he travelled some 7000 miles through at least seventeen different states. During his voyage, Kikumura gradually accumulated the various components he would use to assemble his bombs--for example, tape and wire in Lexington, Kentucky on March 31; switches and jacks in Huntington, West Virginia on April 1; flashbulbs, batteries, and fire extinguishers in Cheltenham, Pennsylvania on April 10; and so on. On April 11, Kikumura unsuccessfully attempted to sell his car and to rent a different car in Philadelphia. The next day, he was arrested at the Vince Lombardi service area while en route to New York, his three bombs by then fully assembled.

In addition to the bombs and bomb paraphernalia, investigators also found in Kikumura's car a map of New York City. Marked on the map was West 24th Street, between 7th and 8th Avenues, about one-half block from a navy and marine recruiting office located at the corner of 7th Avenue and 24th Street. Investigators also found an American Airlines timetable with the handwritten notation, "Friday morning, 330.00, 4/15." Detective Fuentes speculated that Kikumura intended to detonate his bombs at the recruiting office and to board a flight costing $330 on the morning of April 15.

Kikumura's bombs were described in detail by FBI Special Agent James Thurman, one of the country's leading experts on that subject. In order to construct each of his three bombs, Kikumura had sawed the top off of a fire extinguisher, hollowed out its contents, and filled the extinguisher with about three pounds of gunpowder, some wadding, about three pounds of lead shot, and a flashbulb connected to some wire running out of the top of the extinguisher. The bombs would detonate when an electric current passing down the wire caused the bulb to flash. The heat from the flash would then burn the gunpowder, producing a build-up of pressure in the extinguisher and resulting in an explosion that would spray lead shot and metal fragments from the fire extinguisher in all directions. For one of his three bombs, Kikumura had assembled a fusing system out of a timer, a toggle switch, some batteries, and two jack connectors. The system was capable of detonating the bomb up to one hour after Kikumura had started it. Kikumura's car contained materials from which similar fusing systems for each of his other two bombs could have been constructed.

The FBI laboratory report described the gunpowder in Kikumura's bombs as capable of producing a "low explosive main charge" and the lead shot as "enhanc[ing] the fragmentation effect of these devices." According to the report, the lead shot would "produce additional personal injuries and property damage when the devices exploded."

Thurman testified that in a large room with a ten-foot ceiling, Kikumura's three bombs would be capable of producing a fireball extending out to the ceiling and walls. He then explained, however, that the bombs' main destructive force stemmed from Kikumura's use of lead shot as shrapnel:

[T]he real damage from something like this is not from the fireball but the shrapnel that's contained inside the device and the device itself. Here we have two sources of fragmentation. One is from the [the fire extinguisher] itself, [which] would fragment in the explosion. Secondly, the lead shot that is contained within [the fire extinguisher] would be spread around the room, travelling at a very high velocity ... [so as] to injure and possibly kill people within the room. It's impossible to say that everyone in here would be injured or killed. But certainly there would be numerous casualties within the room, as well as ... significant property damage.

On direct examination, Thurman offered his expert opinion that Kikumura placed lead shot in the bombs in order "to injure or kill people." On cross-examination, he conceded that "one of the purposes for this shrapnel could have been to destroy property." On redirect, however, Thurman reiterated that "when lead shot is placed into a device, it's there primarily to injure people, and in the process it will damage property." Furthermore, he agreed that "the reason that it is primarily intended to injure people is [that] it's going to disperse in all different directions, and if anyone is in the room, [the shot will] likely penetrate their [sic] skin." On recross, Thurman agreed that "the spray pattern of the fragments would cause damage to, say, nearby vehicles if that was your intention, or windows, or something like that." Finally, in response to a question from the court, Thurman likened Kikumura's bombs in certain respects to Claymore mines, military devices that project a number of small metal balls in a spray pattern. Like Claymore mines, Thurman opined, Kikumura's bombs were capable of destroying property, but were "meant as anti-personnel devices."

Thurman also explained the significance of three other chemicals found in Kikumura's car when he was arrested--aluminum powder, ammonium nitrate, and mercury. Kikumura possessed about two pounds of aluminum powder, which he had purchased in Nashville, Tennessee on March 30. That powder, combined with the proper amount of ammonium nitrate, could have produced a 50-pound bomb capable of blowing out the doors and bringing down the ceiling of the room in which it was exploded, producing "mass casualties and destruction" in the process. Although Kikumura's car contained only traces of ammonium nitrate, not enough to assemble such a bomb, ammonium nitrate is ordinarily sold in large quantities, so that "if you find traces of ammonium nitrate ... in someone's automobile, that's an indication that the person ... likely had a much larger quantity." Kikumura obtained his mercury by breaking and emptying eleven thermometers he had purchased in Fairmont, West Virginia during or about the first week in April. Thurman testified that he could imagine only two uses for that mercury. First, it might be converted into mercury fulminate, a very powerful explosive often used as a detonator, although Kikumura was not found to possess all of the chemicals required to effect the necessary chemical reaction. Second, the mercury itself might be used as a booby trap for a bomb, so that anyone who tried to move the bomb would set it off.

Thurman testified that the construction of Kikumura's bombs and fusing system indicates a "resourceful individual" with "considerable knowledge" of explosives and "a very high level of expertise."

The government also introduced evidence regarding Kikumura's prior terrorist activities. Jacob Van Mastright, a Dutch police officer, submitted an affidavit stating that he had arrested Kikumura on May 1, 1986, at Schiphol Airport in Amsterdam. The Dutch authorities discovered that Kikumura was attempting to smuggle over two pounds of TNT flakes into the Netherlands inside a cardboard orange drink container hidden in his luggage. Agent Thurman testified that that amount of TNT was capable of producing "mass casualties and fragmentation injuries within [a] room, with the likely possibility of blowing out the doors and the ceiling falling down." Moreover, the Dutch police discovered six detonators hidden inside a transistor radio that Kikumura was carrying. The detonators were concealed so skillfully that the police could not "identify anything particularly wrong with the contents of the radio" even after carefully examining an x-ray of it. Kikumura was detained for about four months. When the airport search was determined to be illegal under Dutch law and no further evidence against Kikumura was discovered, he was deported.

Finally, the district court admitted into evidence, over Kikumura's objection, an affidavit by FBI agent Michael Hartman, who oversees the Bureau's investigation of the JRA. In his affidavit, Hartman detailed at some length his conversations with a confidential informant who had reported to the FBI details of activities inside a terrorist training camp in the Bekaa Valley in Lebanon.

The informant recounted that three members of the JRA arrived at the camp for training sometime during 1984. In January of 1985, two of the three returned for follow-up training, which included, among other things, practice in manufacturing explosives from ammonium nitrate and aluminum powder, and in setting off mercury fulminate detonators with flashbulbs.

During the spring of 1985, a JRA member later identified as Junzo Okudaira spent a week at the camp conducting research on remote control devices. Soon thereafter, the JRA arranged to move into the camp permanently. According to the informant, among the members of the JRA who took up residence there were Okudaira and, in the fall of 1986, Kikumura. While in the camp, Kikumura trained other JRA members. Kikumura admitted to the informant that he had travelled widely and had been arrested once previously.

The JRA leader in the camp, Mariam Shigenobu, told the informant that the JRA's primary enemy was the United States and that the JRA planned to strike within the United States. In June 1987, Shigenobu accepted an invitation to train in Libya.

In his affidavit, Agent Hartman stated that Okudaira is the prime suspect in the bombing of an American U.S.O. club in Naples, Italy, which killed five people and wounded eighteen. That bombing occurred on April 14, 1988, two days after Kikumura was arrested in New Jersey and exactly two years after the United States bombed targets in Libya in retaliation for a previous terrorist bombing in West Berlin. Italian authorities have photographic and fingerprint identifications of Okudaira.

3. The District Court's Findings and Conclusions

The district court made extensive factfinding, describing in often graphic terms the heinous crimes it concluded Kikumura was about to commit. The gist of the district court's findings are summarized in the following introductory paragraph:

In summary, Kikumura is a member of the Japanese Red Army ("JRA"), a violent terrorist organization; he has trained other members of the JRA and received training from the JRA; he has engaged in a course of terrorist conduct in Holland concerning explosives which caused his arrest; he meticulously planned, schemed and attempted to execute a terrorist mission in this country with remarkable skill and deliberation; and he planned to kill and seriously injure scores of people. But for the alert and professional conduct of Trooper Cieplensky, Kikumura would have succeeded in murdering and maiming countless numbers of people--for no reason other than they are Americans.

United States v. Kikumura, 706 F. Supp. 331, 335 (D.N.J.1989).10 In particular, the district court found that Kikumura's bombs "were intended and designed for flesh and blood, not bricks and mortar." Id. at 338.

On the basis of this record and these findings, the district court determined that an upward departure was appropriate. The court relied upon the following factors to justify its substantial departure: intent to cause multiple deaths and serious injuries (see Guidelines Secs. 5K2.1 and 5K2.2); planned detonation of unusually dangerous explosives (see id. Sec. 5K2.6); intent to disrupt a governmental function (see id. Sec. 5K2.7); extreme conduct (see id. Sec. 5K2.8); endangerment of public health/safety and national security (see id. Sec. 5K2.14); and inadequacy of criminal history category (see id. Sec. 4A1.3). 706 F.Supp. at 340-42. In view of these factors, which it did not quantify, and the four purposes of sentencing set forth in 18 U.S.C. Sec. 3553(a)(2)(A)-(D), the court concluded that a sentence of 30 years' imprisonment was reasonable and imposed it. Id. at 346. Kikumura appealed. We have jurisdiction pursuant to 18 U.S.C. Sec. 3742(a)(3).

B. Guideline Departures--General Principles and the Applicable Standard of Proof

The general framework for reviewing departures from the sentencing guidelines is by now settled. A district court must sentence within the applicable guideline range "unless the court finds that there exists an aggravating or mitigating circumstance of a kind, or to a degree, not adequately taken into consideration by the Sentencing Commission in formulating the guidelines that should result in a sentence different from that described." 18 U.S.C. Sec. 3553(b).11 "This provision is mandatory." United States v. Uca, 867 F.2d 783, 786 (3d Cir.1989). Thus, if the circumstances relied upon by the district court to justify the departure were adequately considered by the Sentencing Commission, then its decision to depart must be reversed. See 18 U.S.C. Sec. 3742(f)(1). Our scope of review of this question is plenary. See United States v. Ryan, 866 F.2d 604, 610 (3d Cir.1989). Of course, the circumstances relied upon must in fact exist in the case under consideration, but we are obliged to "accept the findings of fact of the district court unless they are clearly erroneous." 18 U.S.C. Sec. 3742(e).

Once we conclude that a departure is not prohibited by Sec. 3553(b), we must still determine whether the sentence imposed was reasonable, "having regard for--(A) the factors to be considered in imposing a sentence, as set forth in [18 U.S.C. Sec. 3553(a)12 ]; and (B) the reasons for the imposition of the particular sentence, as stated by the district court pursuant to the provisions of section 3553(c)13." 18 U.S.C. Sec. 3742(e)(3). This determination "involves at least two subissues: whether the factors relied on are appropriate; and whether the degree of departure was appropriate." Ryan, 866 F.2d at 610. At this stage of the inquiry, our review is deferential. See id.

Broadly speaking, Kikumura challenges the district court's decision to depart from the guidelines on three different grounds. First, he challenges as clearly erroneous the critical finding that he intended to use his bombs to kill people. Second, he contends that a departure predicated upon his offense conduct is legally prohibited, because the guidelines under which he was sentenced adequately considered his felonious intent. Third, he argues that even if some departure were permissible, the degree of departure in this case was unreasonable. We consider each of these contentions in turn.

Perhaps like no case ever before reported, this one illustrates both the utility of, and the dangers in, real offense sentencing--a system that metes out punishment on the basis of a defendant's actual conduct in a particular case. Such a system recognizes that "particular crimes may be committed in different ways, which in the past have made, and still should make, an important difference in terms of the punishment imposed." Breyer, The Federal Sentencing Guidelines and the Key Compromises Upon Which They Rest, 17 Hofstra L.Rev. 1, 9 (1988). Because criminal statutes have never been (and probably never could be) written with sufficient particularity to take all such factors into account, a system of pure charge offense sentencing--one that metes out punishment solely on the basis of the offense of conviction--would necessarily abstract away considerations obviously relevant in determining an appropriate sentence. For example, it is self-evident that an internationally trained terrorist who is bent on murdering scores of innocent civilians should be sentenced far more severely than a duly licensed explosives merchant who knows that one of his customers intends to blow up an abandoned warehouse in order to commit insurance fraud, even if each of these defendants is convicted under 18 U.S.C. Sec. 844(d) for transporting explosives "with the knowledge or intent that [they] will be used to kill ... any individual or unlawfully to ... destroy any building."

The Sentencing Commission recognized the necessity of some measure of real offense sentencing during each stage of consideration of offense-related conduct.14 First, by writing offense guidelines in general descriptive terms, so that "some statutes may be covered by more than one guideline," Guidelines Sec. 1B1.1 application note 3, the Commission injected an element of real offense sentencing into even the initial determination of some defendants' base offense levels. Second, by modifying all base offense levels through a host of specific offense characteristics, see id. Sec. 1B1.1(b), and adjustments, see id. Sec. 1B1.1(c), (e), it incorporated several real offense factors into the determination of the final offense level of any particular defendant. Third, by designating only a few offense-related factors not otherwise mentioned in the guidelines as irrelevant for sentencing purposes, it consciously chose not to foreclose departure on the basis of most factors not expressly taken into consideration by the Guidelines' specific offense characteristics and adjustments. See id. Ch. 1, Pt. A, intro., 4(b); see also Breyer, supra, at 14.15

Though long recognized as a practical necessity, real offense sentencing can create the potential for significant unfairness. This is so because every factual consideration deemed relevant for sentencing purposes must be established through a collateral, post-verdict adjudication at which the applicable procedural protections are significantly lower than those applicable at the trial itself. For example, a criminal defendant enjoys the right to a trial by jury, see U.S. Const. amend. VI, but that right does not exist at sentencing, see, e.g., Spaziano v. Florida, 468 U.S. 447, 459, 104 S. Ct. 3154, 3161, 82 L. Ed. 2d 340 (1984). At trial, an element of a charged offense must be proven beyond a reasonable doubt, see In re Winship, 397 U.S. 358, 364, 90 S. Ct. 1068, 1072, 25 L. Ed. 2d 368 (1970), but most pertinent sentencing facts need only be established by a preponderance of evidence, see McMillan v. Pennsylvania, 477 U.S. 79, 91, 106 S. Ct. 2411, 2418, 91 L. Ed. 2d 67 (1986) (holding that the preponderance standard is generally constitutional); United States v. McDowell, 888 F.2d 285, 290-91 (3d Cir.1989) (holding that the preponderance standard is generally appropriate in guidelines sentencing). At trial, a jury may consider evidence only if it is admissible under the Federal Rules of Evidence, and the confrontation clause bars consideration of all admissible hearsay unsupported by either "a firmly rooted hearsay exception" or other "particularized guarantees of trustworthiness," Ohio v. Roberts, 448 U.S. 56, 66, 100 S. Ct. 2531, 2539, 65 L. Ed. 2d 597 (1980). At sentencing, by contrast, the Federal Rules of Evidence are inapplicable, see Fed.R.Evid. 1101(d)(3); Williams v. New York, 337 U.S. 241, 250-51, 69 S. Ct. 1079, 1084-85, 93 L. Ed. 1337 (1949) (endorsing the use of hearsay evidence at sentencing), and hearsay normally may be considered subject only to the modest due process requirement that it bear "some minimal indicium of reliability beyond mere allegation," United States v. Baylin, 696 F.2d 1030, 1040 (3d Cir.1982); see also 18 U.S.C. Sec. 3661 ("No limitation shall be placed on the information concerning the background, character, and conduct of a person convicted of an offense which a court of the United States may receive and consider for the purpose of imposing an appropriate sentence."); United States v. Sciarrino, 884 F.2d 95, 97 (3d Cir.) (holding that the reliability analysis is not heightened by the shift from an unstructured sentencing regime to a more determinate guidelines system), cert. denied, --- U.S. ----, 110 S. Ct. 553, 107 L. Ed. 2d 549 (1989).

This dichotomy reflects the judgment that a convicted criminal is entitled to less process than a presumptively innocent criminal defendant, see, e.g., McDowell, 888 F.2d at 290 (noting that the sentencing decision "is rarely ever as crucial as the initial decision finding guilt"), as well as the concern that "over-burdened trial courts would be greatly disserved by the time-consuming hearings" that more intensive procedural protections would require, United States v. Lee, 818 F.2d 1052, 1057 (2d Cir.), cert. denied, 484 U.S. 956, 108 S. Ct. 350, 98 L. Ed. 2d 376 (1987); see also Breyer, supra, at 11 ("[T]he requirement of full blown trial-type post-trial procedures ... would threaten the manageability that the procedures of the criminal justice system were designed to safeguard."). Compare Mathews v. Eldridge, 424 U.S. 319, 334-35, 96 S. Ct. 893, 902-03, 47 L. Ed. 2d 18 (1976), which held as a general matter that the degree of protection required by the due process clause depends, among other things, on balancing the nature of the private interest affected against the government's interest in avoiding the fiscal or administrative burdens entailed by heightened procedural requirements.

In run-of-the-mill sentencing cases, these principles are amply justified. For example, it would be unreasonable to suggest that the determination whether a convicted drug dealer willfully impeded the investigation of his offense, which would increase an applicable sentencing range of between 27 and 33 months to one between 33 and 41 months, see Guidelines Sec. 3C1.1, Ch. 5, Pt. A, must or should be made with the same degree of care as the prior determination whether an accused citizen may at all be stigmatized and punished as a convicted drug dealer.

Because less procedural protection is so clearly appropriate in the majority of sentencing cases, we sometimes tend to regard it as appropriate in all sentencing cases. However, legal rules--even rules that function perfectly well in familiar contexts when stated in categorical terms--cannot always be applied in extreme situations. Obviously, an appropriate level of procedural protection cannot be calibrated on a sliding-scale, case-by-case basis. For example, if proof by a mere preponderance is sufficient to justify a two-level increase for willfully impeding an investigation, see McDowell, 888 F.2d at 288, 290-91, then proof by that identical standard is also appropriate in order to justify, for example, a four-level increase for organizing an offense, see Guidelines Sec. 3B1.1(a); or a six-level increase for unlawfully receiving explosives that one knows to be stolen, see id. Sec. 2K1.3(b)(2); or probably even a ten-level increase for distributing those explosives to a fugitive from justice, see id. Sec. 2K1.3(b)(4). Here, however, we are dealing with findings that would increase Kikumura's sentence from about 30 months to 30 years--the equivalent of a 22-level increase in his offense level, see id. Ch. 5, Pt. A. This is perhaps the most dramatic example imaginable of a sentencing hearing that functions as "a tail which wags the dog of the substantive offense." McMillan, 477 U.S. at 88, 106 S.Ct. at 2417. In this extreme context, we believe, a court cannot reflexively apply the truncated procedures that are perfectly adequate for all of the more mundane, familiar sentencing determinations.

Analytically, there are two possible approaches for providing sufficient process in such situations. The first would place some limit on the concededly broad power of legislatures to define, and courts to consider, conduct that is or could be criminalized as an aggravating factor at sentencing. In effect, this approach would require that, for sentencing purposes, certain findings in certain circumstances be made pursuant to the entire panoply of procedural protections that apply at trial. Neither here nor in the district court, however, did Kikumura advance the argument that a conviction under 18 U.S.C. Sec. 844(d) is too slender a reed on which to support consideration at sentencing of the defendant's specific intent to commit murder. "[I]n the absence of special circumstances, such as a change in the law, we will not consider on appeal an issue that the parties failed to present to the district court." Flick v. Borg-Warner Corp., 892 F.2d 285, 387 (3d Cir.1989). No such special circumstances are present here.

A second, narrower approach to the problem would ratchet up certain, though not necessarily all, of the procedural protections afforded a defendant at sentencing, so as more closely to resemble those afforded at trial. Kikumura expressly advanced such an approach at his sentencing hearing when he argued that in order to support a departure predicated upon his intent to commit murder, the prosecution must establish that intent by clear and convincing evidence. The district court held that a mere preponderance was sufficient, see 706 F. Supp. at 345, but held alternatively that its findings were supported by clear and convincing evidence, see id. at 345 n. 28.

On appeal, neither party raised the standard of proof issue in terms. However, Kikumura has asked us to review findings of fact, an exercise that necessarily entails determining what standard of proof the factfinder should have applied in the first instance. See, e.g., Jackson v. Virginia, 443 U.S. 307, 318, 99 S. Ct. 2781, 2788, 61 L. Ed. 2d 560 (1979) (holding that a habeas court reviewing the sufficiency of evidence underlying a criminal conviction must "determine whether the record evidence could reasonably support a finding of guilt beyond a reasonable doubt " (emphasis added)); cf. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 252, 106 S. Ct. 2505, 2512, 91 L. Ed. 2d 202 (1986) ("[T]he inquiry involved in a ruling on a motion for summary judgment or for a directed verdict necessarily implicates the substantive evidentiary standard of proof that would apply at the trial on the merits.").

McMillan held that a preponderance standard was generally constitutional but suggested that a different question would be presented if the magnitude of a contemplated departure is sufficiently great that the sentencing hearing can fairly be characterized as "a tail which wags the dog of the substantive offense," 477 U.S. at 88, 106 S. Ct. at 2417. For the reasons explained above, we hold that in such situations, the factfinding underlying that departure must be established at least by clear and convincing evidence. Because Kikumura requested no higher standard of proof in his sentencing memorandum, we assume without deciding that the clear and convincing standard is sufficient in such cases. Our position is perfectly consistent with McDowell, which held that a preponderance standard is sufficient to justify adjustments provided for in Chapter 3 of the guidelines, see 888 F.2d at 291, none of which involves more than a four-level increase or decrease to a defendant's offense level.16

We recognize that there is overwhelming authority in our sister circuits for the proposition that guideline sentencing factors need only be proven by a preponderance of evidence, see, e.g., United States v. Wilson, 900 F.2d 1350, 1354 (9th Cir.1990); United States v. Frederick, 897 F.2d 490, 493 (10th Cir.1990); United States v. Guerra, 888 F.2d 247, 250-51 (2d Cir.1989), cert. denied, --- U.S. ----, 110 S. Ct. 1833, 108 L. Ed. 2d 961 (1990); United States v. Ehret, 885 F.2d 441, 444 (8th Cir.1989), cert. denied, --- U.S. ----, 110 S. Ct. 879, 107 L. Ed. 2d 962 (1990); United States v. Urrego-Linares, 879 F.2d 1234, 1237-38 (4th Cir.), cert. denied, --- U.S. ----, 110 S. Ct. 346, 107 L. Ed. 2d 334 (1989); United States v. Wright, 873 F.2d 437, 441 (1st Cir.1989), but we note that in none of these cases did the operative facts involve anything remotely resembling a twelve-fold, 330-month departure from the median of an applicable guideline range. We hold that the clear and convincing standard is, under these circumstances, implicit in the statutory requirement that a sentencing court "find" certain considerations in order to justify a departure, 18 U.S.C. Sec. 3553(b), and we reserve judgment on the question whether it is also implicit in the due process clause itself.

C. Consideration of the Hearsay Statements of a Confidential Informant

In finding by clear and convincing evidence that Kikumura intended to kill people, the district court relied in part on the Hartman affidavit, which described the terroristic intent of the JRA generally and which linked Kikumura to the JRA specifically. The Hartman affidavit consists of hearsay (the declarations of a confidential informant) within hearsay (as related by Agent Hartman). Therefore, Kikumura contends, "[t]he Hartman affidavit should not have been admitted into evidence." Appellant's Br. at 10, 11 n. 3.17

In Baylin, we held that "as a matter of due process, factual matters may be considered as a basis for sentence only if they have some minimal indicium of reliability beyond mere allegation." 696 F.2d at 1040. In response to Kikumura's challenge, the district court faithfully applied Baylin and concluded that that standard was easily met. See 706 F. Supp. at 342-45. We agree. However, we believe that the Baylin standard, like the preponderance standard, is simply inadequate in situations as extreme and unusual as this one. Normally, hearsay statements may be considered at sentencing as long as they comply with the due process requirement set forth in Baylin. But here the district court's reliance upon the out-of-court declarations of a confidential informant was a factor in its decision to depart over ten-fold from the applicable guideline range. When the magnitude of the departure is so disproportionate, we believe that the sentencing court must ratchet up not only the standard of proof, but also the standard of admissibility.

At trial, the confrontation clause imposes a higher constitutional standard of admissibility than does Baylin at sentencing. Absent a firmly rooted evidence rule sanctioning its use, hearsay may be considered at trial only upon a showing of "particularized guarantees of trustworthiness," together with a showing that the out-of-court declarant is unavailable to testify.18 Roberts, 448 U.S. at 66, 100 S.Ct. at 2539. We are, however, reluctant to superimpose the jurisprudence of the confrontation clause upon the sentencing phase. As a textual matter, the sixth amendment, which refers to "criminal prosecutions," arguably applies only at trial. Further, as we discussed earlier, see supra at 1099-1100, the procedural protections afforded a convicted defendant at sentencing are traditionally less stringent than the protections afforded a presumptively innocent defendant at trial. Finally, if we held, as a general matter, that confrontation clause analysis is appropriate at sentencing as well as at trial, we would, in effect, render Baylin a nullity.19

Nonetheless, we believe that the Baylin standard is not sufficiently exacting to be applied in cases, such as this one, where the sentencing hearing can fairly be characterized as a "tail which wags the dog of the substantive offense," McMillan, 477 U.S. at 88, 106 S.Ct. at 2417. In such a situation, we think that due process requires more than a "minimal indicium of reliability," Baylin, 696 F.2d at 1040 (emphasis added).20 Instead, the court should examine the totality of the circumstances, including other corroborating evidence, and determine whether the hearsay declarations are reasonably trustworthy.21 Application of this heightened Baylin test is parallel to the requirement of clear and convincing evidence in the standard of proof context. At a garden variety sentencing hearing, Baylin and a preponderance standard apply; whereas, at trial, Roberts and a reasonable doubt standard apply. But, at a sentencing hearing where the court departs upwards dramatically from the applicable guideline range, intermediate standards should apply. Accordingly, factual findings must be supported by clear and convincing evidence,22 and hearsay statements cannot be considered unless other evidence indicates that they are reasonably trustworthy.

We are satisfied that the Hartman affidavit complies with our new intermediate standard as well as the more forgiving Baylin standard.23 The informant's testimony regarding Kikumura's presence and activities in the terrorist training camp in Lebanon is, in large degree, verified by numerous other details in the record: the presence of ammonium nitrate, aluminum powder, mercury fulminate, and flash bulbs in Kikumura's car; the fact that Kikumura was arrested for similar terrorist conduct in the Netherlands in 1986; the close parallels in conduct between Kikumura in the United States and Okudaira in Naples; the fact that Kikumura purchased the components for his bombs in several states in order to avoid detection; and, perhaps most significantly, the fact that Kikumura was obviously skilled and experienced at constructing potentially lethal, carefully disguised firebombs from materials that could be purchased with ease at a neighborhood hardware store. Kikumura points to nothing in the record that casts doubt upon the believability of the informant's testimony. We conclude that there is sufficient corroboration to establish that the informant's hearsay statements are reasonably trustworthy, and, therefore, the district court's consideration of the Hartman affidavit was proper.

D. Validity of the District Court's Factfinding

Kikumura contends that the district court clearly erred insofar as it found that he intended to use his bombs against people, as opposed to property. He argues that "the testimony of Thurman established only that the lead shot would be useful to enhance personal injury or to enhance property damage, depending upon the bomber's intent. But [one] could not infer which of these two goals a bomber intended by the fact that he used lead shot." Appellant's Br. at 15 (emphasis in original). Kikumura misapprehends the significance of Thurman's testimony. Three times--once on direct, once on redirect, and once in response to a question from the bench--Thurman stated unequivocally that he believed Kikumura's bombs, by virtue of their use of lead shot, were designed for use against people. See supra at 1096. Kikumura points out that Thurman also agreed that "one of the purposes for this shrapnel could have been to destroy property," but that concession does not undermine his unequivocal conclusion that the purpose of the lead shot in fact was to kill and injure people. Contrary to Kikumura's suggestion, the other evidence in the record is, to some extent, probative of Kikumura's intended use of the bombs, and we think that it only reinforces Thurman's opinion.

In crediting Thurman's uncontradicted testimony and concluding that Kikumura's use of lead shot indicates an intent to blow up people by clear and convincing evidence, the district court did not commit clear error.

E. Was an Offense-Related Departure Legally Permissible?

"In determining whether a circumstance was adequately taken into consideration [so as to preclude departure], the court shall consider only the sentencing guidelines, policy statements, and official commentary of the Sentencing Commission." 18 U.S.C. Sec. 3553(b). The offense guidelines that applied to Kikumura's various counts of conviction were Secs. 2K1.3, 2K1.6, 2K2.1, 2K2.2, 2L2.2, and 2L2.4. Kikumura does not contend that Secs. 2K1.3, 2L2.2, or 2L2.4 take into consideration the intent to kill people. He does argue, however, that Secs. 2K1.6, 2K2.1, and 2K2.2 take this intention into consideration.

Kikumura concedes that his prior arrest in the Netherlands would support an offender-related departure under Guidelines Sec. 4A1.3, which authorizes a departure "[i]f reliable information indicates that the criminal history category does not adequately reflect the seriousness of the defendant's past criminal conduct or the likelihood that the defendant will commit other crimes." Section 4A1.3 also provides that "[i]n considering a departure under this provision, the Commission intends that the court use, as a reference, the guideline range for a defendant with a higher or lower criminal history category, as applicable." Plainly, an offender-related departure under Sec. 4A1.3 could not by itself support the sentence imposed in this case. Even if Kikumura's past conduct in the Netherlands and Lebanon indicates that his criminal history is as serious as the criminal histories of most defendants classified as category VI, the highest category employed in the guidelines, that conclusion alone would support a departure only into a range of between 57 and 71 months, see Guidelines Ch. 5, Pt. A, a far cry from the 360-month sentence actually imposed. Accordingly, we must determine whether any offense-related departures are permitted.24

On the date of Kikumura's sentencing, Sec. 2K1.6 read as follows:

Sec. 2K1.6 Shipping, Transporting, or Receiving Explosives with Felonious Intent or Knowledge; Using or Carrying Explosives in Certain Crimes

(a) Base Offense Level (Apply the greater):

(1) 18; or

(2) If the defendant committed the offense with intent to commit another offense against a person or property, apply Sec. 2X1.1 (Attempt or Conspiracy) in respect to such other offense.25

Application of Sec. 2K1.6 was predicated upon Kikumura's conviction under 18 U.S.C. Sec. 844(d), which makes it unlawful to transport any explosive in interstate commerce "with the knowledge or intent that it will be used to kill, injure, or intimidate any individual or unlawfully to damage or destroy any building, vehicle, or other real or personal property." Kikumura's offense level of 18 was calculated by application of Sec. 2K1.6(a)(1).

Kikumura contends that "18 U.S.C. Sec. 844(d) has intent to destroy property or to kill or injure people as a material element of the offense." Appellant's Br. at 17 (emphasis in original); see also id. at 20 ("[T]he statute presumes that the 'average' Sec. 844(d) violator intends to unlawfully blow something up (including people)...."). Therefore, Kikumura reasons, the guideline to which that statute is referenced, Sec. 2K1.6, necessarily takes into account his specific intent to kill people.

Kikumura glosses over the import of the precise statutory language. The statute's usage of the passive voice--one must know or intend that one's explosives "will be used" to harm people or damage property--brings within its ambit two distinct classes of offenders: those who transport explosives themselves intending to harm people or damage property, and those who transport explosives with intent or knowledge that others will or might use them to harm people or damage property. We think that Sec. 2K1.6 distinguishes between these two classes of Sec. 844(d) violators and, not surprisingly, that it punishes the former more harshly than the latter. In our view, if "the defendant" himself intended to kill people, then under Sec. 2K1.6(a)(2), he may be sentenced as if he had been convicted of attempted murder, for which the base offense level is 20, see Guidelines Sec. 2A2.1.26 On the other hand, if a defendant transports explosives knowing that they "will be used" by someone else to kill people, but does not himself specifically intend to kill people, he may be sentenced only under Sec. 2K1.6(a)(1), which carries a base offense level of 18.

Contrary to Kikumura's argument, the " 'average' Sec. 844(d) violator" does not necessarily intend to blow up something or someone himself. See supra at 1099. In fact, the "average" Sec. 844(d) violator sentenced at level 18 under Sec. 2K1.6(a)(1) does not specifically intend to kill people himself, because the "average" Sec. 844(d) violator who does is sentenced at no less than level 20, pursuant to Sec. 2K1.6(a)(2). We therefore find no merit in Kikumura's argument that the Sentencing Commission took into account a defendant's specific intent to kill people when it fixed 18 as an appropriate offense level for those Sec. 844(d) violators to be sentenced under Sec. 2K1.6(a)(1).27

Kikumura also attempts to prove his point by contrasting Sec. 2K1.6 with Sec. 2K1.3, which provides a base offense level of only 6 for unlawfully transporting explosives. Kikumura is correct to the extent he contends that Sec. 2K1.6 metes out significantly higher punishment for transporting explosives illegally than does Sec. 2K1.3 precisely because Sec. 2K1.6 considers the transporter's felonious intent regarding the ultimate use of those explosives. The fact that Sec. 2K1.6, as a whole, adequately considers felonious intent is beside the point, however, for the issue before us is the far more precise question whether Sec. 2K1.6(a)(1) adequately considers a defendant's specific intent to kill people himself. For the reasons already stated, we conclude that it does not.

Finally, Kikumura contends that a departure in these circumstances is prohibited by United States v. Uca, 867 F.2d 783 (3d Cir.1989). We disagree. Uca involved two defendants convicted of attempting to procure, without a license, a large number of handguns for sale to overseas purchasers, in violation of 18 U.S.C. Secs. 922(a)(1)(A) and 922(a)(3). The defendants were sentenced under Guidelines Sec. 2K2.3. The district court departed upwards in part because it could not imagine any legitimate purpose for which the buyers would use the handguns. See 867 F.2d at 786. In language seized upon by Kikumura, we reversed, holding that "[t]he intended use of the guns is not a circumstance warranting upward departure." Id. at 790.

Obviously, since Uca addressed what the Commission did or did not adequately consider in writing Sec. 2K2.3, it does not control our determination of what the Commission did or did not consider in writing Sec. 2K1.6. This distinction aside, we still do not find Uca 's reasoning applicable to this case.

As discussed above, we think that, in most cases, there is a meaningful distinction between a defendant who intends to cause a harm himself and a defendant who acts with the knowledge or intent that someone else cause the same harm. Moreover, each of these two mental states can be further trifurcated according to whether the harm was specifically intended, knowingly caused, or consciously and unjustifiably risked. Compare, e.g., Model Penal Code Sec. 2.02(2)(a) (1985) ("Purposely") with id. Sec. 2.02(2)(b) ("Knowingly") with id. Sec. 2.02(2)(c) ("Recklessly"). Even though these various distinctions obviously blend into one another at the margins, we think it obvious that the most culpable of the six possible mental states (acting oneself with specific intent to cause a harm) is significantly more culpable than the least culpable (acting with reckless disregard for whether someone else will cause the same harm). We believe that this case involves the former, and that Uca involved only the latter.

In relevant part, Sec. 2K2.3, at the time Uca was decided, provided as follows:

Sec. 2K2.3 Prohibited Transactions in or Shipment of Firearms and Other Weapons

(a) Base Offense Level

(1) 12, if convicted under 26 U.S.C. Sec. 5861; or

(2) 6, otherwise.

....

(c) Cross Reference

(1) If the defendant provided the firearm to another for the purpose of committing another offense, or knowing that he planned to use it in committing another offense, apply Sec. 2X1.1 (Attempt or Conspiracy) in respect to such other offense, if the resulting offense level is higher.28

So far as appears in our opinion, the Uca defendants did not themselves intend to use any of the firearms to commit a crime. Moreover, they could not have specifically intended or known that their buyers would use the purchased firearms to commit any crime, for if they did, their offense level would have been calculated under Sec. 2K2.3(c)(1), not under Sec. 2K2.3(a)(2), as it was, see 867 F.2d at 785.

The district court in Uca cited no hard evidence that the defendants knew, at least in any particularized sense, that their buyers would commit crimes with the handguns. Rather, the court relied on a kind of res ipsa loquitur argument that handguns--especially handguns purchased from unlicensed distributors--are often used in connection with other criminal activity, together with the inference that the defendants must have been aware of this fact, but were indifferent to the "potential acts of violence" that their conduct would facilitate. See 867 F.2d at 786. The district court's reasoning on this score might have been sound, but it failed to distinguish defendants' conduct from that of most other defendants sentenced under old Sec. 2K2.3(a)(2) for selling firearms without a license. Accordingly, we reversed the district court's upward departure.

In sum, we do not find Uca on point, despite our imprecise language in that case about the irrelevance of a defendant's "unlawful purpose," 867 F.2d at 788, or of the "intended use" of unlawfully transported firearms, id. at 790. Uca involved an entirely different guideline and an entirely different mental state from those at issue here. Even in the context of the guideline it was construing, Uca plainly cannot, in light of the cross-reference to Sec. 2X1.1 contained in Sec. 2K2.3(c)(1), stand for the proposition that Sec. 2K2.3(a)(2) adequately considers a defendant's specific intent regarding the ultimate unlawful use of firearms, any more than Sec. 2K1.6(a)(1) can be read, in light of Sec. 2K1.6(a)(2), as adequately considering a defendant's specific intent to kill people.

Kikumura next contends that both of the firearms provisions applicable to him, Guidelines Secs. 2K2.1 and 2K2.2, adequately take into consideration his intent to kill people. At the time of Kikumura's sentencing, those sections punished, respectively, possession of firearms by certain prohibited persons and possession of firearms in violation of various regulatory provisions of the National Firearms Act, 26 U.S.C. Sec. 5801 et seq.29 In pertinent part, they provided:

Sec. 2K2.1 Receipt, Possession, or Transportation of Firearms and Other Weapons by Prohibited Persons

(a) Base Offense Level: 9

(b) Specific Offense Characteristics

....

(2) If the defendant obtained or possessed the firearm solely for sport or recreation, decrease by 4 levels.

(c) Cross Reference

(1) If the defendant used the firearm in committing or attempting another offense, apply the guideline in respect to such other offense, or Sec. 2X1.1 (Attempt or Conspiracy) if the resulting offense level is higher than that determined above.

Sec. 2K2.2 Receipt, Possession, or Transportation of Firearms and Other Weapons in Violation of National Firearm Act

(a) Base Offense Level: 12

(b) Specific Offense Characteristics

....

(3) If the defendant obtained or possessed the firearm solely for sport, recreation or collection, decrease by 6 levels.

(c) Cross Reference

(1) If the defendant used the firearm in committing or attempting another offense, apply the guideline for such other offense or Sec. 2X1.1 (Attempt or Conspiracy), if the resulting offense level is higher than that determined above.

For his one count of conviction under 18 U.S.C. Sec. 922(g)(5), Kikumura's offense level was determined to be 9, pursuant to Sec. 2K2.1(a). For each of his six counts of conviction under either 26 U.S.C. Sec. 5861(d) or (i), Kikumura's offense level was determined to be 12, pursuant to Sec. 2K2.2(a).

Strangely enough, Kikumura's argument in the context of Secs. 2K2.1 and 2K2.2 is less easily dismissed than his Sec. 2K1.6 argument, even though Sec. 2K1.6, as a general matter, punishes conduct more serious than the conduct punished under the firearms possession guidelines. As explained above, Kikumura's argument that Sec. 2K1.6(a)(1) adequately takes into consideration a defendant's specific intent to kill people can be rejected simply by examining the face of Sec. 2K1.6(a)(2). Here, however, the cross-references to Sec. 2X1.1 are not dispositive, because they become applicable only after at least an attempt (to commit murder), see Guidelines Secs. 2K2.1(c)(1) and 2K2.2(c)(1), not just upon the presence of mere intent (to commit murder), see id. Sec. 2K1.6(a)(2). Under well-settled principles of the law of attempts, Kikumura's conduct in this case amounted only to "mere preparation," not an attempt, despite the extraordinarily meticulous nature of that preparation and the strong degree of certainty regarding his ultimate intent. See, e.g., Model Penal Code Sec. 5.01 (1985).

Nonetheless, we believe that anomalous results would follow were the firearms possession guidelines construed so as to preclude consideration of an intent to commit murder as a permissible basis for departure whenever the cross-references to Sec. 2X1.1 are inapplicable. To see this, one need only contrast a hypothetical operation of the firearms possession guidelines with a hypothetical operation of the firearms transaction guideline, discussed above in connection with Uca. Imagine the following two situations:

1. D unlawfully possesses a gun and intends to use it to kill V himself. D begins to prepare for the murder, but is apprehended just before the preparation ripens into an attempt.

2. D lawfully possesses a gun, but gives it to T intending that T use it to kill V. T begins to prepare for the murder, but is apprehended just before the preparation ripens into an attempt.

We think it self-evident that defendant D in case 1 is at least as culpable as defendant D in case 2, if not more so. In case 2, D would be sentenced at no less than level 20, as if convicted for attempted murder, pursuant to Sec. 2K2.3(c)(1) of the old firearms transfer guideline. See Guidelines Sec. 2K2.3(c)(1) (cross-referencing attempt guideline); id. Sec. 2A2.1 (providing a base offense level of 20 for attempted murder). On Kikumura's reading of the firearms possession guidelines, however, because "[t]he guidelines begin with a presumption that illegally possessed weapons are intended for unlawful use against people, and that presumption is factored into the base offense level," Appellant's Br. at 17-18, in case 1, D could be sentenced at no greater than level 12 were his unlawful possession of the firearm within the ambit of Sec. 2K2.2, and at no greater than level 9 were his unlawful possession within the ambit of Sec. 2K2.1. Because the Sentencing Commission obviously intended for the guidelines to operate as a cohesive and integrated whole, we have a duty to construe them in pari materia. We highly doubt that the Commission intended such a strange result as this one, and it would take clear textual evidence in the guidelines to persuade us otherwise.

Kikumura contends that Secs. 2K2.1(b)(2) and 2K2.2(b)(3) provide that evidence. We disagree. As the application notes make clear, the Commission intended these provisions as downward adjustments for "intended lawful use." Guidelines Sec. 2K2.1 application note 1 (Oct. 1988 ed.); id. Sec. 2K2.2 application note 1. Thus, we agree with Kikumura to the extent he suggests that the base offense levels incorporate some presumption of intended unlawful use. That use, however, might be nothing more than an unauthorized resale at a profit, so as to trigger potential application of Sec. 2K2.3. Moreover, many typical intended unlawful uses of a firearm--for example, to commit a robbery--do not imply an altogether more grave intent to discharge the firearm at the victim of the robbery. We desire to avoid the anomalous result described above, and we are mindful of a general interpretive posture under which the Commission's silence regarding a particular circumstance should ordinarily be construed as an intent not to foreclose departure, see Guidelines Ch. 1, Pt. A, intro., 4(b). We therefore conclude that these various possible intended unlawful uses of a firearm are sufficiently different from the intended unlawful use of shooting and killing someone that the consideration of intended unlawful uses generally in the base offense levels in Secs. 2K2.1(a) and 2K2.2(a) does not preclude departure on the basis of an intent to commit murder. We reject Kikumura's argument accordingly.

We hold that none of the offense guidelines under which Kikumura was sentenced adequately takes into consideration Kikumura's intent to commit murder. Thus, we conclude that an upward departure was legally permissible under 18 U.S.C. Sec. 3553(b).30F. Was the Departure Reasonable?

Having determined that an upward departure was not prohibited by Sec. 3553(b), we must next decide whether the sentence imposed was in fact reasonable. See 18 U.S.C. Sec. 3742(e)(3).

We begin by considering "whether the factors relied on [by the district court] are appropriate" bases for departure. Ryan, 866 F.2d at 610. It is important to note that analytically this inquiry is distinct from the prior inquiry into whether departure based on the same factors is foreclosed by Sec. 3553(b). For example, the guidelines nowhere take into consideration the fact that a defendant has blue eyes, but obviously any departure predicated upon that consideration would be unreasonable. In this case, we have had no trouble concluding that a defendant's intent to commit murder, when not taken into consideration under any of the applicable guidelines, is an eminently reasonable basis for an upward departure.

We must now consider a more difficult question--"whether the degree of departure was reasonable." Ryan, 866 F.2d at 610. Congress's express statutory directive to the courts of appeals to affirm all departures that are not unreasonable, see 18 U.S.C. Sec. 3742(f)(3), makes clear that our scope of review is deferential. At this stage, the question is no longer whether the district court has substituted its judgment for that of the Sentencing Commission, but rather whether the court of appeals should substitute its judgment for that of the district court. As one court has cogently stated:

This [final] step involves what is quintessentially a judgment call. District courts are in the front lines, sentencing flesh-and-blood defendants. The dynamics of the situation may be difficult to gauge from the antiseptic nature of a sterile paper record. Therefore, appellate review must occur with full awareness of, and respect for, the trier's superior "feel" for the case.

United States v. Diaz-Villafane, 874 F.2d 43, 49-50 (1st Cir.), cert. denied, --- U.S. ----, 110 S. Ct. 177, 107 L. Ed. 2d 133 (1989). For these reasons, we have held that when departures are not foreclosed by Sec. 3553(b), the district courts are entitled to "a substantial amount of discretion" in determining the extent of any departure. See Ryan, 866 F.2d at 610.

At the same time, however, we think it clear that there must be some objective standards to guide the determination of reasonableness. The substantial sentencing disparity against which the Sentencing Reform Act was a reaction resulted not because district judges are generally unreasonable, but because determining an appropriate sentence for a particular defendant is quintessentially the kind of inquiry about which reasonable people may differ. Reasonable people might also differ when determining the limits of reasonableness of a departure, but we must make that determination as a matter of law. Reviewing the degree of departure is especially difficult because there are no statutory or common law calipers to guide us. Under the old regime, sentencing discretion was essentially unreviewable. In contrast, under the aegis of the Sentencing Reform Act, we think it improper for us to measure the extent of a departure from the guidelines against only our hunches and instincts. Rather, we believe that fidelity to the policy undergirding the guidelines requires us at least to strive for some principled basis for reviewing the reasonableness of departures. As the Seventh Circuit suggested in United States v. Ferra, 900 F.2d 1057, 1062 (7th Cir.1990), "[u]nless there is discipline in determining the amount of departure, ... sentencing disparity will reappear."

A natural starting point for deriving such standards is the statutory provision governing appellate review of sentences, 18 U.S.C. Sec. 3742. It provides that in determining whether the amount of a departure was unreasonable, the court of appeals must consider: "(A) the factors to be considered [by the district court] in imposing a sentence, as set forth in [18 U.S.C. Sec. 3553(a) ]; and (B) the reasons for the imposition of the particular sentence, as stated by the district court pursuant to the provisions of section 3553(c)." Id. Sec. 3742(e)(3). Unfortunately, neither of these considerations provides much guidance. Subsection 3742(e)(3)(B) seems to say little more than that if a district court fails to state the specific findings on which its departure is based, the court of appeals should vacate and remand for clarification, rather than attempting to reason from the record evidence to the sentence imposed in the first instance.

Section 3553(a) provides a standard more verbose, but no more precise, than the reasonableness requirement. It directs the district court to impose a sentence "sufficient, but not greater than necessary, to comply with the purposes set forth in paragraph (2) of this subsection." Paragraph (2), in turn, lists four purposes of sentencing--namely, the need for the sentence imposed

(A) to reflect the seriousness of the offense, to promote respect for the law, and to provide just punishment for the offense;

(B) to afford adequate deterrence to criminal conduct;

(C) to protect the public from further crimes of the defendant; and

(D) to provide the defendant with needed educational or vocational training, medical care, or other correctional treatment in the most effective manner.

Id. Sec. 3553(a)(2). Put more succinctly, the statute requires the sentence imposed to be minimally sufficient to satisfy concerns of retribution, general deterrence, specific deterrence, and rehabilitation--the four major theories that undergird criminal law. The statute says nothing about how much of each factor the law requires, how to determine the amount of each factor a contemplated sentence in fact would provide, or how to weigh factors against one another in the frequent situations when different factors (for example, retribution and rehabilitation) tend to pull in opposing directions.

The Sec. 3553(a) criterion is so vague that even though the Sentencing Commission has publicly encouraged departures in cases where application of its concededly imperfect guidelines would result in an inappropriately high or low sentence, see Guidelines Ch. 1, Pt. A, intro., 4(b), courts have held in effect that a refusal to depart can never result in a sentence inconsistent with Sec. 3553(a). See, e.g., United States v. Colon, 884 F.2d 1550, 1553 (2d Cir.), cert. denied, --- U.S. ----, 110 S. Ct. 553, 107 L. Ed. 2d 550 (1989); see also United States v. Denardi, 892 F.2d 269, 272 (3d Cir.1989) (endorsing Colon ). In short, we do not think that Sec. 3553(a), by itself, could possibly impose enough "discipline in determining the amount of departure," Ferra, 900 F.2d at 1062, to prevent unwarranted disparity from continuing to plague the large class of cases not controlled by Sec. 3553(b).

Recognizing the need for additional standards, the courts of appeals have recently begun to look to the guidelines themselves for guidance in determining the reasonableness of a departure. See, e.g., United States v. Landry, 903 F.2d 334, 340-41 (5th Cir.1990); United States v. Pearson, 900 F.2d 1357, 1362 (9th Cir.1990); Ferra, 900 F.2d at 1061-63; United States v. Kim, 896 F.2d 678, 684-85 (2d Cir.1990). Today, we endorse that general approach.

The guidelines themselves prescribe such an approach with respect to offender-related departures pursuant to Sec. 4A1.3. If a defendant's criminal history category seems an inadequate measure of "the seriousness of the defendant's past criminal conduct or the likelihood that the defendant will commit other crimes," Guidelines Sec. 4A1.3, the district court is authorized to depart. In this context, however, the power to depart is not the power to " 'throw away the guidelines,' " Ferra, 900 F.2d at 1061-62, in favor of any sentence that strikes the judge as reasonable. Rather, Sec. 4A1.3 provides that, "[i]n considering a departure under this provision, ... the court [should] use, as a reference, the guideline range for a defendant with a higher or lower criminal history category, as applicable." For example, if a category I offender has committed a level 18 offense, and if the actual seriousness of the defendant's criminal history is like those of most defendants classified as category III offenders, then the district court must impose a sentence of no higher than 41 months--the top of the range applicable to level 18 offenses committed by category III offenders. See Guidelines Sec. 4A1.3.31

A similar approach is required by statute when a district court imposes sentence for an offense with no applicable guideline. In that context, the court must have "due regard for the relationship of the sentence imposed to sentences prescribed by guidelines applicable to similar offenses and offenders." 18 U.S.C. Sec. 3553(b). For example, if a category I offender commits offense-related conduct most analogous to conduct that the guidelines classify as level 18, then the district court must impose a sentence of no higher than 33 months--the range applicable to level 18 offenses committed by category I offenders.

The principle employed in each of these contexts--that the appropriate length of a sentence should be determined from the sentencing table, even outside the context of straightforward applications of underlying offense and offender guidelines--extends naturally into the context of offense-related departures pursuant to Guidelines Sec. 5K2.0.32 For example, if the guidelines applicable to a category I offender indicate an offense level of 18, but fail to take into consideration certain aspects of his offense-related conduct, a departure is authorized. However, if the actual seriousness of the defendant's offense-related conduct is like that of conduct classified at level 20, then, we believe, the district court must impose a sentence of no higher than 41 months--the top of the range applicable to level 20 offenses committed by category I offenders.

In our view, analogy to the guidelines is also a useful and appropriate tool for determining what offense level a defendant's conduct most closely approximates. Often, a departure might be supported by aggravating conduct that itself would constitute a separate offense described under a different guideline. In such cases, the reasonableness of a departure may be evaluated by "treat[ing] the aggravating factor as a separate crime and ask[ing] how the defendant would be treated if convicted of it." Ferra, 900 F.2d at 1062. That exercise necessarily would involve application of the Guidelines' grouping rules, Guidelines Secs. 3D1.1 to .5, which ensure that as new offenses are considered, total punishment is increased, if at all, in decreasing marginal increments. In these cases, analogy to the guidelines is especially appropriate because "[i]t would throw the structure of the guidelines out of kilter to say that a defendant may receive more time on a 'departure' than he could have received had he been convicted of the crimes leading the judge to depart." Ferra, 900 F.2d at 1063.

Sometimes, the aggravating circumstance not adequately taken into consideration will amount not to a separate crime, but to a special offense characteristic employed in offense guidelines other than those applied to the defendant.33 In such cases, the gravity attached to the characteristic in the other guidelines provides appropriate guidance as to what degree of departure would be reasonable. For example, several guidelines provide for a two-level increase in offense level if the defendant's conduct involved "more than minimal planning." See, e.g., Guidelines Secs. 2A2.1(b)(1), 2B1.1(b)(4), 2F1.1(b)(2). If, in a case involving guidelines that do not take more than minimal planning into account, a district court decided to depart upwards solely on the basis of a defendant's more than minimal planning, we think that a departure equivalent to increasing the defendant's offense level by more than two levels would be presumptively unreasonable.

Thus, in a case involving a level 18 offense committed by a category I offender, a more than minimal planning departure would be reasonable only if the sentence imposed did not exceed 41 months--the high end of the guideline range applicable to level 20 offenses committed by category I offenders. In order to justify a departure of a greater magnitude than that, the district court would have to explain why more than minimal planning should make a greater difference in the context before it than it makes under the various guidelines where it is deemed a specific offense characteristic, or else explain why the case involves aggravating factors other than a defendant's more than minimal planning. Cf. United States v. Shuman, 902 F.2d 873, 876-77 (11th Cir.1990) (holding that the degree of a departure based on conduct described as a specific offense characteristic in inapplicable guidelines was reasonable when the departure amounted to an increase in offense level no greater than that provided in the inapplicable guidelines).

We do not suggest that this approach for measuring the reasonableness of a departure can always be mechanically applied, in the way that the guidelines can be mechanically applied to determine the applicable sentencing range from which possible departures may be considered. We are dealing here with analogies to the guidelines, which are necessarily more open-textured than applications of the guidelines. Sometimes, the district court will want to depart on the basis of conduct best conceived of as a separate offense for which there is no applicable guideline. Other times, the relevant conduct will appear most like a specific offense characteristic or an adjustment that the guidelines nowhere employ. Still other times, competing analogies each will be plausible, or a series of factors will be present that cannot be neatly disentangled from one another and analyzed component by component. In short, "the reasonableness of a departure does not always depend on ... mathematical precision." Shuman, 902 F.2d at 877. However, by attempting to "link the extent of departure to the structure of the guidelines," Ferra, 900 F.2d at 1062, the courts can avoid the kind of standardless determinations of reasonableness that inevitably produce unwanted disparity.

There may be vehicles for making offense-related departures under Ch. 5, Pt. K of the guidelines (and for determining the reasonableness thereof) other than the kind of analogic reasoning outlined above. Acknowledging the wide variety of situations that may arise and the superior vantage point of the district courts, we do not rule out the development of other such vehicles. We also recognize that there may be cases where the guidelines will not afford useful analogies. Under the facts of this case, however, we know of and have been directed to no other means to assess the reasonableness of the district court's departures; hence, we apply analogic reasoning. In applying this methodology, the deference we owe to the district courts, see Ryan, 866 F.2d at 610, attaches to the details of its exercise--i.e., how to craft any particular analogy it might wish to employ, which in turn could affect, for example, the determination whether offense conduct classified under applicable guidelines at level 18 is better treated, pursuant to a departure, as the equivalent of level 19, or 20, or some other plausible level.

We turn now to the application of this analysis to Kikumura. Initially, we note that the district court, without the benefit of the principles we announce today, measured the degree of its departure only against the standard of Sec. 3553(a). Ordinarily, we would simply vacate forthwith and remand for application of the analogic reasoning outlined above. We have examined the district court's opinion with care, however, and we are convinced beyond any doubt that the district court would impose as high a sentence as possible up to 30 years.34 If a reasonable analogy existed to support the sentence imposed, remand would be a pointless exercise.35 We therefore proceed to consider whether such analogy exists. In this inquiry, we feel bound to consider only those circumstances cited by the district court as warranting a departure. Cf. 18 U.S.C. Sec. 3553(c); id. Sec. 3742(e)(3)(B). Given the principles enunciated at length above, we also are bound to defer to the district court to the extent that its departure can be deemed reasonable. Because in the upper ranges of the guidelines, which are applicable here, the length of imprisonment is more significantly affected by the criminal history category, we will first consider the available criminal history departure.

a. Criminal history category

The district court made clear that an upward departure pursuant to Sec. 4A1.3 was appropriate in order to reflect the inadequacy of Kikumura's criminal history category. Section 4A1.3 provides that "[i]f reliable information indicates that the criminal history category does not adequately reflect the seriousness of the defendant's past criminal conduct or the likelihood that the defendant will commit other crimes," the court may depart from the applicable guideline range. Applying this section, the district court found that both Kikumura's 1986 arrest in the Netherlands and his terrorist training in Lebanon justified an upward departure. See 706 F. Supp. at 341.

Kikumura concedes that his actual criminal history could reasonably support an increase from category I to category IV, but he nonetheless contends that the district court, in departing upwards, improperly considered his activities in Lebanon. See Appellant's Br. at 24-25. Regarding Kikumura's conduct in Lebanon, the district court found that he provided training in the use of explosives to members of a group publicly committed to perpetrating acts of terrorism against Americans. This conduct, which was not considered in Kikumura's criminal history category, appears to satisfy both parts of Sec. 4A1.3. First, it might be evidence that Kikumura's criminal history does not adequately reflect the seriousness of his past criminal conduct.36 Second and more importantly, Kikumura's training in Lebanon surely indicates that his criminal history under-represents the likelihood that he will engage in other criminal conduct.37 In sum, the record reflects, by clear and convincing evidence, that Kikumura is a professional terrorist who is extremely likely to commit other equally serious crimes in the future.38 Indeed, it is hard to imagine a situation where recidivism is more likely. Based upon his activities in the Netherlands and Lebanon, we think it would be reasonable to analogize Kikumura to a category VI offender. We now turn to the available offense level departures.

b. Intent to kill people

The district court found that Kikumura intended to use his bombs to kill people. See 706 F. Supp. at 340. That finding, which is amply supported by the record, suggests analogy to the attempted murder guideline. It might be objected that acting upon an intent to commit murder involves conduct far more inchoate than actually completing an attempt to commit murder. The analogy to attempted murder is nonetheless reasonable because several of the explosives and firearms guidelines cross-reference a court to the attempted murder guideline based upon a defendant's intent to kill people. See Guidelines Secs. 2K1.4(c)(1), 2K1.6(a)(2), 2K2.3(c)(1). Attempted murder carries a base offense level of 20. See id. Sec. 2A2.1(a).

Kikumura, however, did not intend to kill just one person; he intended to kill "scores." 706 F. Supp. at 340. As Thurman testified, had Kikumura's bombs been detonated, they would have caused "numerous" casualties. Thus, in further developing its analogy, the district court reasonably might have considered multiple attempted murders, not just one. Through application of the grouping rules, attempting to kill six people, as opposed to one, would increase a defendant's offense level by five. See Guidelines Secs. 3D1.1 to .4.39 Although it is difficult to attach an exact number to the word "scores" on this record, we think it clear that a five-level increase would be justified. One who intends to set off a powerful, time-delayed bomb inside a room full of people, we think, can fairly be said to have intended to kill whoever is present when the bomb explodes.40 The district court's findings make out a compelling circumstantial case that Kikumura intended for his bombs to be detonated in some crowded area. It seems inconceivable to us that someone with Kikumura's background and training travelled to another continent and prepared as he did for over a month in order to commit an isolated act of purely symbolic violence. We think that the district court would be quite reasonable to assume that Kikumura intended to detonate his bombs when at least six people were present. As a result, the district court could reasonably have increased his offense level by five.

An argument could be made that because of the unusual circumstances in this case, and the possibility that scores of people would have been killed, the court could adjust upwards by more than five levels. However, we believe that to allow such an adjustment, in the absence of stronger identification of a target and given the obvious difficulty of accomplishing such a bombing without detection, would be to permit undue speculation. In any event, we do not find clear and convincing evidence on that score, and hence we conclude that the maximum allowable upward adjustment is five levels.

c. More than minimal planning

Using the specific offense characteristics in the attempted murder guideline as a starting point, the district court reasonably might have enhanced Kikumura's sentence on account of his extraordinarily detailed, extended, and meticulous preparation in assembling deadly bombs from unremarkable items purchased throughout the eastern United States. We recognize that the attempted murder guideline contains a more than minimal planning adjustment only in cases involving an actual assault, see id. Sec. 2A2.1(b)(1), thus indicating that the Commission thought more than minimal planning generally irrelevant in cases involving attempts without an actual assault, which were sentenced under the same guideline. Nonetheless, we think that Kikumura's extraordinarily meticulous planning makes this aspect of his conduct substantially more culpable than that of most defendants charged with attempted murder. We therefore conclude that the district court reasonably could have imposed a two-level increase on this ground. See id. Sec. 5K2.0 (authorizing departure if, "in light of unusual circumstances," a specific offense characteristic explicitly mentioned in the applicable guideline attaches an "inadequate" level to that characteristic).

d. Intent to disrupt a governmental function

The district court departed upwards in part because of Kikumura's intent to disrupt a governmental function. See 706 F. Supp. at 342; Guidelines Sec. 5K2.7. Kikumura challenges this departure on two grounds.

First, Kikumura argues that this section, which authorizes departure "[i]f the defendant's conduct resulted in a significant disruption of a governmental function" (emphasis added), is inapplicable because it extends only to completed conduct. In attempting to parse the language of this provision so finely, Kikumura misapprehends the nature of the factors enumerated as possible grounds for departure in chapter 5K2. The Commission spoke in terms of completed conduct because most offenses involve completed conduct, not attempts. But fitting a case within the literal language of a Sec. 5K2 provision is neither a necessary nor a sufficient condition for making an offense-related departure. It is unnecessary because the enumeration of factors in chapter 5K2 is not meant to be exhaustive. See Guidelines Sec. 5K2.0. It is not by itself sufficient because certain guidelines do take into consideration certain of the enumerated factors. See id. In short, the enumerated factors merely flag for the court "some of the factors that the Commission has not been able to fully take into account." Id. (emphasis added). Ultimately, the permissibility of the departure is determined under 18 U.S.C. Sec. 3553(b), not under the literal applicability of some factor listed in chapter 5K2.

Second, Kikumura argues that in any event, his planned terrorist bombing of an uncertain target does not constitute "disruption of a governmental function" within the meaning of the guidelines. We agree. The district court determined that Kikumura sought to disrupt governmental functions because he "intended to alter through illegal means American anti-terrorist policies." 706 F. Supp. at 342. In essence, the court equated intent to influence government decisionmaking with intent to disrupt governmental functions. Even though our standard of review is deferential, we believe that the district court's interpretation of Sec. 5K2.7 exceeds the bounds of reasonableness.

Despite our warning against overly literal interpretations of the factors listed in chapter 5K2, we think that Sec. 5K2.7 was meant to deal with a wholly different situation than was involved here. Section 5K2.7 talks in terms of "disruption" of or "interference" with a governmental "function." As we read this language, the Commission intended to punish more harshly criminal conduct that frustrates the normal, day-to-day operation of government. We do not think that such language is sufficiently elastic to encompass illegal conduct which is designed to change governmental policy. Rather than being overly technical, we believe that this interpretation of Sec. 5K2.7 is simply faithful to the Commission's intention.

The case law is in accord, being to the effect that Sec. 5K2.7 applies only to conduct that directly obstructs or impedes the normal operation of government, not to unlawful efforts to change government policy. See, e.g., United States v. Murillo, 902 F.2d 1169, 1174 (5th Cir.1990) (departure valid under Sec. 5K2.7 because defendant, who was convicted of immigration law violations, compromised "the integrity of the amnesty system"); United States v. Garcia, 900 F.2d 45, 48 (5th Cir.1990) (departure valid where defendant stole an inordinately large volume of mail); United States v. Burns, 893 F.2d 1343, 1347 (D.C.Cir.1990) (departure valid where defendant manipulated federal procurement mechanism in order to divert government resources); United States v. Clemente, 729 F. Supp. 165, 168 n. 2 (D.Mass.1990) (departure would have been valid where defendant was convicted of stealing police promotional exams and selling them to police officers).

We acknowledge that other guidelines list an intent to influence government action as a factor warranting an upward adjustment, see, e.g., Guidelines Sec. 2A4.1(b)(1) (kidnapping); id. Sec. 2Q1.4(b)(4) (tampering with public water). Further, as the Commission has expressly stated, the fact that a factor is listed as a specific offense characteristic in one guideline, but not in others, does not preclude consideration of that factor as a basis for departure from the other guidelines. See Guidelines Sec. 5K2.0. Notwithstanding these considerations, we believe that the Commission's inclusion of this factor in a few guidelines does not mean that it should be reflexively applied to all guidelines. More fundamentally, however, without clear guidance from the Commission, we are reluctant to interpret this portion of the kidnapping and public water guidelines expansively, because we think that the across-the-board application of this factor to all crimes may be problematic in first amendment terms.

According to the district court's factual findings, Kikumura planned to bomb a federal government building in New York City in order to dissuade United States policymakers from engaging in any future anti-terrorist strikes similar to the bombing of Libya. While the vehicle chosen by Kikumura to influence American policy would shock the sensibilities of even the most callous among us, in theory, the "intent to influence" aspect of his conduct is indistinguishable from the motivation underlying ordinary civil disobedience designed to change government policy. Although those who break the law for political reasons are not immune from punishment, it is not clear that they should be punished more severely just because their ultimate objective was to influence government policy. Moreover, to assign a stiffer penalty to conduct with a political component than would be applicable to the identical conduct without a political component is arguably tantamount to putting a price on political activism. This seems to us a matter that the Commission may wish to consider further, and absent clearer instructions from the Commission at this time, we decline to approve an upward departure based on intent to disrupt a governmental function.e. Extreme conduct and endangerment of public safety

The district court also departed upwards in part because of Kikumura's extreme conduct. See 706 F. Supp. at 341; Guidelines Sec. 5K2.8. Section 5K2.8 authorizes an upward departure to reflect the nature of the conduct "[i]f the defendant's conduct was unusually heinous, cruel, brutal, or degrading to the victim," which includes cases of "torture of a victim, gratuitous infliction of injury, or prolonging of pain or humiliation."41 In order to avoid double-counting, we must limit our focus only to whether Kikumura's intended conduct was unusually heinous within the universe of attempted murders, for the attempted murder guideline plainly accounts for the fact that attempted murder, by its very nature, involves heinous conduct. Cf. Guidelines Sec. 5K2.1 (authorizing departure if death results and explaining further that the "means by which life was taken" is relevant in determining the appropriate extent of such a departure).

We conclude that an extreme conduct departure would be justified on these facts. Kikumura's bombs would have operated by spraying shrapnel with tremendous force in every direction. Were they detonated in a crowded room, it is almost a statistical certainty that some people would be killed instantly, others would escape largely unharmed, and still others would sustain permanent injuries or die slowly from their wounds. Because of this last group, the district court reasonably could conclude that an extreme conduct departure from the attempted murder guideline would be appropriate.

Furthermore, Sec. 5K2.14 states that the court may increase a sentence above the guideline range "[i]f national security, public health, or safety was significantly endangered" by the defendant. The district court found that "Kikumura's terrorist plot presented a grave threat to public health and safety" and therefore departed upwards pursuant to this section. 706 F. Supp. at 341. We acknowledge that the extreme conduct and public safety grounds for departure may overlap. We nonetheless believe that these concepts are analytically distinct. On the one hand, the extreme conduct ground addresses the fact that if Kikumura had exploded his bombs, some of his victims would have suffered particularly painful injuries or deaths. On the other hand, the public safety consideration addresses the fact that if Kikumura had detonated his bombs in a dense metropolitan area such as Manhattan, he would have significantly endangered the safety and welfare of the general public. Cf. United States v. Chiarelli, 898 F.2d 373, 380-81 (3d Cir.1990) (upholding a Sec. 5K2.14 departure because defendants' attempted escape at high speed in a van significantly endangered the safety of local residents). This latter ground for departure takes into account that Kikumura's planned bombing of a New York City building was far more reckless, and hence more culpable, than a planned bombing of a building in a more isolated area would have been.

There is also a possibility of overlap between our departure for endangerment of public safety and our consideration of multiple attempted murders. Yet we are still convinced that the public safety ground for departure addresses a sufficiently distinct aspect of Kikumura's conduct to warrant an upward departure. Because estimating the number of Kikumura's potential victims is an inherently speculative endeavor, we concluded that Kikumura could not reasonably be viewed as having attempted to murder more than six people. See supra at 1115-16. Consistent with that determination, the public safety ground addresses the fact that if Kikumura had successfully detonated his bombs, he not only would have killed people, but also would have threatened the well-being of countless other innocent bystanders.

The guidelines offer no guidance as to how much of an extreme conduct or public safety departure is appropriate. However, the guideline adjustments set forth in Chapter Three each provide for an increase of between two and four levels. Considering all the circumstances, we think that the maximum allowable upward departure (i.e., the limit of "reasonableness") for extreme conduct and endangerment of public safety would be five levels. This would bring Kikumura's offense level to 32--that is, 20 (attempted murder) + 5 (multiple attempted murders) + 2 (more than minimal planning) + 5 (extreme conduct and endangerment of public safety).

There are several reasons why we are convinced that it would be unreasonable to depart upwards by more than five levels for extreme conduct and endangerment of public safety. First, at the upper end of the guidelines, where the offense level grid intersects with criminal history category VI, each incremental offense level carries with it an additional sentence of over two years.42 As a result, we do not take lightly the decision to tack on another offense level. Second, the extreme conduct and public safety departures are superimposed on an already very serious baseline--attempted murders.43 Any increase above five levels for extreme conduct and endangerment of public safety therefore might involve us in impermissible double-counting. Third, this is a situation where the offense is sufficiently inchoate that it does not even rise to the level of an attempt. To endorse a greater departure would thus require us to engage in the sort of unconstrained speculation that we are unwilling to do. Finally, we have already taken into account the fact that Kikumura is a professional terrorist, and by definition a probable recidivist, by countenancing the maximum upward criminal history departure allowed under the guidelines. In view of these facts, we cannot affirm as reasonable any upward departure beyond offense level 32.44

f. Summary

Although the actual guidelines applied to Kikumura indicate a level 18 offense committed by a category I offender, based on the facts articulated as departures bases, the district court could reasonably have analogized this case to one involving a level 32 offense committed by a category VI offender. For such an offender, the applicable guideline range is between 210 and 262 months. Under these circumstances, the district court's sentence of 360 months is unreasonable and must be set aside.

For all the foregoing reasons, we will vacate the judgment of sentence and remand this case to the district court with instructions to resentence Kikumura consistent with this opinion.

ROSENN, Circuit Judge, concurring.

I concur in the court's holding that a sentencing judge may not make extreme departures from the sentencing guidelines without abiding by standards of procedural fairness more stringent than those which usually apply in sentencing hearings.1 I also concur in the court's analysis under the guidelines and in the sentence imposed upon Kikumura. I write separately, however, to express my concern that the Government's manipulation of Kikumura's charge and sentencing illustrates the problem reported by many courts that the sentencing guidelines have replaced judicial discretion over sentencing with prosecutorial discretion. In so doing, it may have violated Kikumura's right to due process.

The labyrinthian course which the sentencing guidelines require courts to navigate in reaching their decisions might obscure one salient feature of Kikumura's prosecution and sentencing: Kikumura was not convicted of attempted murder and yet that is the crime upon which the court relied most heavily in sentencing him. The original indictment handed down by the grand jury charged Kikumura in count two with transporting explosives with the "intent that it will be used to kill, injure, and intimidate one or more individuals." Before trial, the Government deliberately deleted that phrase from the indictment, allowing Kikumura to stipulate to the facts and be convicted of the remaining charges, the most serious of which involved Kikumura's knowledge or intent that the bombs would be used to damage property.2 See United States v. Kikumura, 706 F. Supp. 331, 334 (D.N.J.1989). Thus, the court did not convict Kikumura of the intent to kill, injure, and intimidate one or more individuals, an offense of which Kikumura consistently protested he was innocent. The Government then proceeded at sentencing under guideline Sec. 2K1.6(a)(2) which permitted consideration of Kikumura's alleged intent to kill people with his bombs.

This strategy, of course, made the Government's burden of proving such felonious intent much lighter. For example, prior to the sentencing hearing, the Government was not compelled to furnish Kikumura with the evidence against him to assure an informed and able cross-examination, and at the hearing, the Government was not burdened by the Federal Rules of Evidence and the responsibility to prove each fact beyond a reasonable doubt.

The sentencing guidelines no doubt encouraged this strategy. The guidelines permit the sentencing court to consider relevant conduct not charged in the indictment, including evidence of the defendant's intent in committing the offense. See Guidelines Sec. 1B1.3(a)(4), eff. Oct. 1987.3 Indeed, this comports with the former practice of sentencing judges who traditionally have considered collateral information affecting the level of a defendant's culpability. However, because of the extreme departure involved here for the separate offense of attempted murder, it seems evident that the Government and the sentencing judge did not consider Kikumura's attempt to kill as collateral but primary. In such a case, Kikumura should have been charged and tried for that offense. Failure to do so should preclude the Government from relying upon the separate crime of attempted murder as the vehicle for the drastic enhancement of the defendant's sentence.

A number of courts and the recent Federal Courts Study Committee have noted that one unintended effect of the sentencing guidelines has been to replace traditional judicial discretion over sentencing with prosecutorial discretion. In its final report, the Study Committee noted that "we have been told that the rigidity of the guidelines is causing a massive, though unintended, transfer of discretion and authority from the court to the prosecutor. The prosecutor exercises this discretion outside the system." Report of the Federal Courts Study Committee, April 1990 at 138.

In its earlier tentative recommendations, the Committee aptly observed, "Instead of achieving the congressional purpose of limiting and regulating sentencing discretion, the guidelines have actually had the perverse effect of transferring discretion from the court to the prosecutor, who then exercises the discretion outside the system...." Federal Courts Study Committee Tentative Recommendations for Public Comment, December 22, 1989 at 62, cited in U.S. v. Holland, 729 F. Supp. 125, 133-32 (D.D.C.1990). See also U.S. v. Roberts, 726 F. Supp. 1359, 1363 (D.D.C.1989) ("the de facto transfer of much of the responsibility for sentencing from impartial judges to prosecutors has had the effect of disturbing the due process balance essential to the fairness of criminal litigation"); U.S. v. Boshell, 728 F. Supp. 632, 637 (E.D.Wash.1990) ("Congress has thus shifted discretion from persons who have demonstrated essential qualifications to the satisfaction of their peers, ... to persons who may be barely out of law school with scant life experience and whose common sense may be an unproven asset").

Such prosecutorial control, although not unconstitutional per se, if left unchecked by the courts could be applied in a manner which encroaches upon the constitutional rights of defendants. Indeed, district courts have already found such constitutional violations. See U.S. v. Roberts, 726 F. Supp. 1359, 1368 (D.D.C.1989). In Roberts the district judge found a due process violation where the prosecutor regularly transferred certain drug cases pending in D.C. Superior Court to federal court. The court noted,

Simply by selecting out some defendants from among the many in Superior Court for prosecution in this Court, the U.S. Attorney is able to ensure that these particular defendants will be sentenced to imprisonment for a minimum of five or ten years, while others, not so selected but otherwise similarly situated, will, upon conviction, be sentenced under the discretionary sentencing provisions available in Superior Court to a small fraction of that time.

Roberts, 726 F. Supp. at 1372.

I am concerned that the Government's tactic in obtaining Kikumura's long sentence comes into conflict with the Constitution. What the Government appears to have done is to deliberately collateralize at the charge and trial stage the most critical element for this sentencing, Kikumura's specific intent in transporting the explosives. A hypothetical prosecution may illustrate the perceived misconduct. Suppose the police apprehend a man who is driving recklessly with the intention to meet others in a robbery conspiracy. State officials only charge and convict the man with violating traffic ordinances, but at the man's sentencing hearing argue that the underlying motive for the man's speeding was participation in a robbery at another end of town. The sentencing judge finds the state's evidence convincing and sentences the defendant as if he had been convicted of conspiracy to commit a robbery. I believe that in a federal habeas corpus proceeding attacking the constitutionality of the conviction, we would conclude such conduct violated the man's right to have the most serious crime specifically charged and proven at a trial which included the full panoply of procedural protection. Here, the Government's analogous conduct in manipulating the charge and sentence of Kikumura is no less offensive simply because it has occurred within the shadow of the sentencing guidelines.

Because Kikumura himself did not raise these objections on appeal and the Government had no opportunity to respond, I concur rather than dissent. For this reason, I believe that we are precluded from reversing on the ground that the Government's conduct here infringed upon Kikumura's right to due process. However, I believe that if the Government had proceeded properly and introduced the evidence of Kikumura's intent to kill at trial, the record at trial probably would have justified his conviction of that charge and sentencing as determined by this court.