Full opinion text
LYNCH, Circuit Judge. Identity theft is said to be among this country’s fastest growing crimes. In the 1990’s and earlier, Alan Scott, a former paralegal handy with documents, apparently enhanced his income through an extensive array of white collar crimes using the identities of others. Those of his activities that took place in the late 1990’s led to a series of indictments and three separate criminal cases against him, one case a year from 1997 to 1999; all three cases led to convictions. He appeals in each and we handle the three cases in this omnibus opinion. We take the cases in chronological order. In a 1997 case, Scott pled guilty to bank fraud, 18 U.S.C. § 1344 (1994), and to making and possessing a forged check, 18 U.S.C. § 513(a) (1994). Scott took checks from a Boston law firm that had employed him as a legal assistant in 1995. The cheeks were payable to clients and to the firm; Scott deposited the checks, with forged endorsements, in bank accounts in Texas. For these crimes he received a combined sentence of 96 months and was ordered to pay restitution of $1,381.00. With his conditional plea, Scott reserved his right to appeal certain issues: whether venue in Massachusetts was improper as to the charges, and whether his motion to suppress evidence (on the same grounds as in the income tax cases, discussed below) was wrongfully denied. He also appeals the restitution order. That is appeal No. 00-1381. In a 1998 case, a jury convicted Scott of conspiring to make and of making false claims to an agency of the United States, 18 U.S.C. §§ 286, 287 (1994). Scott filed twenty false income tax returns with the IRS for the tax year 1996 seeking tax refunds in the names of at least twelve people. He used four coconspirators to carry out the scheme, and the intended loss to the government exceeded $80,000.00. For these crimes he received sentences of 96 months for the conspiracy and 60 months for the false returns, concurrent with each other and with the sentence imposed in the 1997 case; he was also ordered to pay restitution of $37,970.68. Scott now appeals the district court’s denial of his motion to suppress certain evidence, several of its evidentiary rulings, and some of its sentencing decisions. Those are appeals Nos. 99-2236, 00-1379, and 00-2350. In a 1999 case, Scott also pled guilty to an additional and different bank fraud, 18 U.S.C. § 1344, and to conspiring to commit that fraud, 18 U.S.C. § 371 (1994). The scheme involved fraudulently obtaining bank automobile loans. For this crime he was sentenced to a 46 month sentence, consecutive to those for the earlier cases, and was ordered to pay restitution of $35,500.00. With his conditional plea, Scott reserved the right to appeal the denial of his motion to dismiss for claimed Speedy Trial Act violations and the denial of a motion to suppress evidence. He also appeals the sentence and restitution order. Those are appeals Nos. 00-1767 and 00-1669. Several issues presented by this appeal are novel to us or worth emphasizing, and we list them in the order we discuss them: 1. In analyzing venue, we modify this Circuit’s “key verb” approach in light of recent Supreme Court cases. 2. We decide that reasonable suspicion of the fraud of attempting to pass a bad check sufficient to justify an investigative stop under Terry v. Ohio does not alone amount to a suspicion that the suspect is armed and dangerous sufficient to justify a frisk. 3. We apply the inevitable discovery doctrine to a defendant whose crime would inevitably have been discovered from the statements of a code-fendant where those statements were given before the codefendant received his Miranda warnings. 4. We uphold the admissibility of opinion testimony of a non-expert witness authenticating or identifying handwriting and discuss the relationship between Federal Rules of Evidence 701 and 901(b)(2). 5. We discuss how restitution orders in cases where two or more defendants are ordered to pay restitution for the same loss should be handled to be clear that the restitution required does not exceed the sum of the loss. 6. We hold on the facts of this case that a trial court is not in compliance with the Speedy Trial Act when it takes a motion to suppress under advisement for over 120 days, disposes of almost all matters, requests additional filings as to certain materials, and relies alone on this request for additional filings to say there was compliance with the Act. We turn to the eases. I. The 1997 Bank Fraud and Forged Securities Case No. 00-1381 Scott pled guilty to this offense. His conditional plea reserved the issues we discuss. A summary of the facts follows. From April to December of 1995, Scott was employed as a paralegal at a small Boston area law firm. Knowing that Scott had a criminal record, a partner of the firm wanted to give him an opportunity to get his life straight. In July of that year Scott opened by mail two accounts at the USAA Federal Savings Bank in San Antonio, Texas. He soon mailed deposits to those accounts using funds that were not his. The law firm regularly received checks payable to its clients (or to its clients and the firm) from insurance companies. Scott stole five of these checks in or -around September and October of 1995. In October he deposited all five of them, with forged endorsements, into his Texas bank accounts. The stolen checks were apparently sent by mail (or the private equivalent) to the bank. Unsurprisingly, the postmarks were not kept. Scott himself appears to have forged at least one of the endorsements on the checks. At the time the checks were deposited, Scott was working at the firm and living in Massachusetts under supervised release. Under the terms of his supervised release, Scott was not to leave Massachusetts. After Scott was arrested in Natick, Massachusetts on December 5, 1995 — we discuss the arrest, relevant to the later cases, in Part II of this opinion — a partner of the firm visited Scott and became convinced Scott had taken the checks. A. Venue Scott says that venue for each of the forgery and bank fraud offenses was properly in Texas, not Massachusetts, and so his convictions should be invalidated. We review legal conclusions de novo, Campos-Orrego v. Rivera, 175 F.3d 89, 96 (1st Cir.1999), and give deference to the district court’s factual conclusions as to venue, United States v. Ochoa, 229 F.3d 631, 636 (7th Cir.2000). The right to be tried in the appropriate venue is one of the constitutional protections provided to defendants by the Sixth Amendment. The burden of showing proper venue is on the government, which must do so by a preponderance of the evidence. United States v. Lanoue, 137 F.3d 656, 661 (1st Cir.1998). We review the evidence on venue in the light most favorable to the government. Id. Two recent Supreme Court opinions on venue cause us to shift somewhat our prior approach to venue questions. In 1993 this Circuit endorsed the use of the “key verb” approach. United States v. Georgacarakos, 988 F.2d 1289, 1293 (1st Cir.1993). That approach analyzed the key verbs in the statute defining the criminal offense in order to determine the scope of the relevant conduct. Id.; United States v. Tedesco, 635 F.2d 902, 905 (1st Cir.1980). To the extent there was any suggestion in Georgacarakos that the “key verb” approach was an exclusive approach, as opposed to one helpful approach, that suggestion is now abandoned. In 1999, the Supreme Court said that words other than the verbs used in a statutory definition of an offense also have import for venue: [W]e have never before held, and decline to do so here, that verbs are the sole consideration in identifying the conduct that constitutes an offense. While the “verb test” certainly has value as an interpretive tool, it cannot be applied rigidly, to the exclusion of other relevant statutory language. The test unduly limits the inquiry into the nature of the offense and thereby creates a danger that certain conduct prohibited by statute will be missed. United States v. Rodriguez-Moreno, 526 U.S. 275, 280, 119 S.Ct. 1239, 143 L.Ed.2d 388 (1999). Rather, the Court, as it had earlier done in United States v. Cabrales, 524 U.S. 1, 6-7, 118 S.Ct. 1772, 141 L.Ed.2d 1 (1998), used a three part test for venue. Rodriguez-Moreno, 526 U.S. at 279-81, 119 S.Ct. 1239. Venue must be determined from the nature of the crime alleged, determined by analyzing the conduct constituting the offense, and the location (or, if the crime is a continuing one, locations) of the commission of the criminal acts. If the crime consists of distinct parts, taking place in different localities, then venue is proper wherever any part can be proved to have taken place. Id. That is the test we now apply. The nature of the crime of forgery is determined by the statutory definition, which provides: Whoever makes, utters or possesses a counterfeited security of a State or political subdivision thereof or of an organization, or whoever makes, utters or possesses a forged security of a State or political subdivision thereof or of an organization, with intent to deceive another person, organization, or government shall be fined under this title or imprisoned for not more than ten years, or both. 18 U.S.C. § 513(a) (1994). The language of the bank fraud statute, in turn, provides: Whoever knowingly executes, or attempts to execute, a scheme or artifice— (1) to defraud a financial institution; or (2) to obtain any of the moneys, funds, credits, assets, securities, or other property owned by, or under the custody or control of, a financial insti-' tution, by means of false or fraudulent pretenses, representations, or promises; shall be fined not more than $1,000,000 or imprisoned not more than 30 years, or both. 18 U.S.C. § 1344 (1994). Scott’s argument is that there was no evidence he forged any endorsements in Massachusetts or mailed any checks from Massachusetts. He argues that the checks could have been mailed from anywhere. Indeed, he supplied an affidavit that he was a frequent traveler during this period, often away from Massachusetts, although the terms of his supervised release forbade exactly such travel. Scott then relies on the supposed absence of facts and invokes the hoary common law presumption that forged checks are presumed to be made where they are first found in their altered state or uttered. See United States v. Britton, 24 F. Cas. 1239, 1241 (C.C.D.Mass.1822) (No. 14,650). Since it is undisputed the checks were “uttered” in Texas, he argues, venue belongs in Texas. The argument is flawed. The presumption operates, as Justice Story said in Britton, only when there is no evidence of where the instrument was uttered, forged, or possessed. Id. That is not this case, as we show. The district court considered that the theft of the checks was the first step in “making” the forgery. It relied on the Second Circuit’s decision in United States v. Delia, 944 F.2d 1010 (2d Cir.1991). Scott rejoins that Delia should be rejected, that taking the checks was merely a preparatory step and that, under this Circuit’s law in Georgacarakos, preparatory steps are not relevant to a venue analysis. Scott’s argument overreads Georgacara-kos, but the Delia issue of whether theft of the instrument is part of “making” the forgery is academic here and we need not pursue it. Other facts establish the necessary connection. It is rational to infer that soon after Scott stole the checks in Boston, the checks were fraudulently endorsed and mailed. Scott endorsed at least one of the checks, and the checks, stolen between September and October, were all deposited in October. This makes it highly likely that both activities — the endorsements and the mailings — took place in Massachusetts. That is, it is highly likely that Scott both “made” and “possessed” the forged checks, see 18 U.S.C. § 513(a), in Massachusetts. Speed was essential to Scott’s forgery scheme. As soon as the law firm discovered the checks were gone, the firm would have notified the insurance companies, which would have sought to stop payment on the checks. Furthermore, that Scott was on supervised release at the time also suggests the need for speed as part of the scheme. Finally, Scott says he acted alone in this scheme, which further reinforces the likelihood he performed these acts in Massachusetts. The bank fraud analysis has a different cast. One of the elements of bank fraud is proof of a scheme to defraud. United States v. Brandon, 17 F.3d 409, 424 (1st Cir.1994). Taking the checks was an act necessary to this particular scheme. For this reason and those given before, there was adequate evidence the offense was both begun and continued in Massachusetts. Bank fraud is also a continuing offense under 18 U.S.C. § 3237(a) and so venue was proper wherever the offense was “begun, continued, or completed.” Scott also makes a constitutional policy argument that the purposes of the Sixth Amendment venue guarantee are offended by finding venue in Massachusetts. The argument fails. This is not an instance of government forum shopping or of a venue with the barest connection to the defendant. This is not at all like Cabrales, in which the Supreme Court upheld a dismissal for improper venue. 524 U.S. at 3-4, 118 S.Ct. 1772. There, the defendant money launderers’ only connection with the Missouri venue was that the money, laundered entirely in Florida, was derived from illegal drug sales in Missouri. Id. The district court’s venue determination was correct. B. Restitution Scott also challenges the district court’s restitution order. He claims there is a dispute as to the proper amount of restitution, and that the district court’s restitution determination was unsupported by a preponderance of the evidence. The district court sentenced Scott to pay $1,381.00 in restitution to the victim insurance companies, unless he had already reimbursed them. There is no dispute as to the amount of the uttered checks. Scott’s claim is simply that the award of restitution is unjustified because, he says, the insurance companies were reimbursed and therefore suffered no actual loss. But the district court’s restitution order was contingent on whether Scott had already reimbursed the insurance companies, and there is no evidence that he did. Accordingly, there is no issue as to the restitution order. II. The 1998 False Income Tax Returns Case Nos. 99-2236, 00-1379, and 00-2350 Scott makes three categories of challenges to the trial court’s rulings, raising nine separate issues. He challenges the court’s denial of his motion to suppress the fruits of the searches of his residence and automobile. He challenges certain eviden-tiary rulings, primarily the decision to permit an IRS agent to give an opinion as to whether handwriting on a number of documents was Scott’s own. Finally, he challenges two elements of the punishment: the enhancement of his sentence on the ground that he was a leader or organizer of the fraudulent scheme, and the restitution order. A brief overview of the facts is helpful. In 1996, the IRS received a set of at least twenty fraudulent tax returns, each requesting a refund of several thousand dollars. These returns used genuine names and Social Security numbers, but the taxpayers whose names and numbers appeared on the returns did not file the returns; they listed genuine employers, but the taxpayers did not work for those employers. The salaries and refund requests were wholly fabricated. Key similarities between the returns — identical names, employers, wages, expenses, and other entries — pointed to a common origin. Some of the returns were deposited directly to an account at the USAA Federal Savings Bank in San Antonio, Texas; others were made out to various addresses in the Boston area. The IRS became aware of the fraud and began an investigation. Details on the false returns led agents to suspect Scott, whom they already knew from past investigations and who had a substantial criminal record. For example, some of the returns listed Turner Construction — a real corporation unrelated to the fraud — as an employer, and Scott had used Turner Construction’s name in some of his earlier frauds. In addition, Scott had an account in his own name at the USAA Federal Savings Bank that came to the IRS’s attention in an earlier search, although that account was not directly used in the tax fraud scheme. As a result, IRS agents began surveillance of Scott in October 1997. The investigating agents obtained search warrants for Scott’s residence in Jamaica Plain, Massachusetts, and for his car in November 1997. They executed the warrants on November 25 and found evidence in Scott’s bedroom linking Scott to the false income tax returns. This evidence included identification cards, materials for making identification cards, a mailbox key, various handwritten notes of names and addresses, and other documents. At trial, Scott moved to suppress the results of the 1997 search based in part on the 1997 warrant’s reliance on several earlier searches that occurred in 1989, 1992, and 1995. To decide this case, we need describe only the events leading to the 1995 search. These events occurred on December 5, 1995, when Scott apparently drove Brian Stephens, a codefendant in this case, to a Circuit City store in Natick, Massachusetts. There, Stephens attempted to pay for a camcorder by cheek. Stephens identified himself as Thomas Judge to the store’s employees and presented false identification bearing that name, an address, and a date of birth. The employee to whom he presented the check submitted it to a routine electronic verification service, which returned an unfavorable result. Stephens then left the store without reclaiming his identification. The store’s manager, David Homsi, became suspicious and watched Stephens as he left. Homsi noted that Stephens left in a white Pontiac Bonneville driven by a white male, wrote down the license number of the car, and called the Natick police. Officer Daniel Brogan, who responded to the call, told Homsi to call the police again if Judge returned. At that time, Homsi gave Brogan the identification card. Shortly thereafter, Stephens called the store, asking permission to retrieve his identification. Stephens then returned and encountered Brogan and another officer. Brogan observed that Stephens matched both the description given by Homsi and the picture on the Judge identification card. On seeing the police, Stephens began to run towards Scott’s car, which was parked nearby. Stephens was stopped at the end of the row where Scott was parked. The officers ordered Stephens to lie down and, after he obeyed, placed him in the back of a police cruiser. Brogan then approached Scott’s car, which Homsi indicated to be the car that Stephens had used earlier. The car was parked away from the entrance to the nearest store, in a relatively secluded area. Its motor was running. Brogan walked up to the car and began to question Scott, asking him why he was waiting in the parking lot. Scott replied that he was waiting for a call and showed Brogan a beeper. Brogan then asked Scott if he knew the man (Stephens) whom the police had just detained in the parking lot. Scott answered that he did not. Brogan concluded that Scott was lying and that he was involved with Stephens in a crime. Brogan advised Scott of his rights, and ordered him out of the car. He pat frisked Scott and found nothing. Brogan then searched the passenger compartment of the car. The search was not merely visual; he opened the glove compartment. In the glove compartment, he found a hypodermic needle, possession of which without a prescription or other justification violated state law, Mass. Gen. Laws ch. 94C, § 27(a) (2000); for this offense, he arrested Scott. At the suppression hearing, Brogan testified that he searched the car because it was cold out and he did not want to put Scott back into the car without making certain it contained no weapons. Brogan’s police cruiser was nearby and could have been used for questioning. Brogan then returned to Stephens (whom he still knew only as Judge) and together with another officer began to question him. Brogan asked Stephens his age and place of residence; Stephens gave answers different from those on the false Thomas Judge identification card. When Brogan asked where Stephens had obtained the information on the card, Stephens replied that he had taken it from a telephone book. Brogan then arrested Stephens for attempting to pass a bad check and for conspiring with Scott to do so. An inventory search of Scott’s car revealed some of the materials Scott sought to have suppressed in this trial, which included employee identification cards and a birth certificate. These materials linked Scott to Ralph Swoboda, in whose name a false tax return relevant to this case was filed. A further link to Lee Morrison, a coconspirator in this case, was established by a Social Security card bearing Morrison’s name; the Natick police found the card in Scott’s wallet in the course of booking. They also found in the wallet a deposit slip and a check from an account of Scott’s at the USAA Federal Savings Bank — the same bank, although not the same account, to which the conspirators in this case arranged for some of the fraudulently obtained tax refunds to be sent. A. Denial of the Motion to Suppress Our review of the ultimate determinations of probable cause and reasonable suspicion on a motion to suppress is de novo. Ornelas v. United States, 517 U.S. 690, 699, 116 S.Ct. 1657, 134 L.Ed.2d 911 (1996). Review of subsidiary factual findings is for clear error. Id. Law enforcement officers have probable cause for a search when “the facts and circumstances within their ... knowledge and of which they had reasonably trustworthy information [are] sufficient in themselves to warrant a man of reasonable caution in the belief that” the search will reveal the fruits or instrumentalities of a crime. Brinegar v. United States, 338 U.S. 160, 175-76, 69 S.Ct. 1302, 93 L.Ed. 1879 (1949) (quoting Carroll v. United States, 267 U.S. 132, 162, 45 S.Ct. 280, 69 L.Ed. 543 (1925)) (bracketed alteration in original) (internal quotation marks omitted); see also Ornelas, 517 U.S. at 696, 116 S.Ct. 1657. Two warrants issued in 1997 for the search of Scott’s home and his car. An affidavit in support of the warrants contained references to material found in seven earlier searches, including searches in 1989, 1992, and 1995. Scott makes a wholesale attack on all of the warrants, but much of the attack is a sideshow. For purposes of this case, Scott needed primarily to suppress the results of the 1997 searches, although some evidence from the 1995 search was also introduced against him at trial. The 1989 and 1992 searches are pertinent here only if the 1997 warrants depended on the results of each of those searches for the requisite probable cause showing. See United States v. Veillette, 778 F.2d 899, 904 (1st Cir.1985) (“[T]he [illegally obtained evidence] should be set to one side ... and the remaining content of the affidavit examined to determine whether there was probable cause to search, apart from the tainted aver-ments.”)- We conclude that the 1989 and 1992 searches are plainly unnecessary to support probable cause for the 1997 searches and therefore unnecessary to the resolution of this case. The 1995 search results were, however, much more important to this case. As the affidavit used to apply for the 1997 warrants states, the 1995 search produced evidence “that directly links [Scott] to the questionable refund scheme” that led to this case. This evidence may have been crucial to the application for the 1997 warrants, and some of it was introduced directly against Scott at trial. Moreover, the government agrees that the validity of the 1995 search is material to the 1997 conviction, No. 00-1381, which we discuss in Part I, and so we consider whether that search was adequately justified. Our conclusion is that the 1995 search was not subject to the exclusionary rule, so no arguable issue is presented as to the use of the results of that search in the 1997 warrant. As such, the 1997 warrant has more than adequate support, and the district court was correct to deny the motion to suppress. 1. The 1995 Warrantless Search of Scott’s Car a. Constitutionality of the Search Law enforcement officials gained access to the contents of Scott’s car via a routine inventory search after the Natick police impounded the car incident to Scott’s arrest on charges of possessing illegally a hypodermic needle. If the arrest was valid, the inventory search was constitutional even though the police had neither a warrant nor probable cause. Colorado v. Bertine, 479 U.S. 367, 371, 107 S.Ct. 738, 93 L.Ed.2d 739 (1987). Whether the arrest was valid depends on the constitutionality of Brogan’s actions leading up to the arrest. Our inquiry focuses first on Brogan’s temporary detention of Scott, which we examine as a seizure under the Fourth Amendment, and second on Brogan’s order that Scott step out of the car and the subsequent sweep of the car, which we examine as searches. If any of these measures was unconstitutional, that illegality may taint the subsequent arrest and inventory search. See Wong Sun v. United States, 371 U.S. 471, 484-87, 83 S.Ct. 407, 9 L.Ed.2d 441 (1963). We assume based on the district court’s opinion that Scott was not free to leave, or at least reasonably believed he was not free, when Brogan began questioning him. The government has argued on appeal that the record would support a conclusion that Scott might have left at any time until Brogan ordered him out of the car. Were this so — were Brogan’s actions in questioning Scott fully consensual — we would engage in no Fourth Amendment scrutiny of the questioning, as a suspect’s voluntary conversation with police is neither a search nor a seizure under the Fourth Amendment. Florida v. Bostick, 501 U.S. 429, 434, 111 S.Ct. 2382, 115 L.Ed.2d 389 (1991). The district court made no finding on whether Brogan exercised his authority to detain Scott, choosing instead to analyze the reasonableness of Brogan’s actions under Terry v. Ohio, 392 U.S. 1, 88 S.Ct. 1868, 20 L.Ed.2d 889 (1968) — and thus at least implicitly assumed that Scott did not consent to converse with Brogan. We follow and approve this part of its analysis. The Fourth Amendment rule laid down in Terry permits the police to detain temporarily a suspect on a reasonable suspicion, supported by articulable facts, that the suspect has committed or is about to commit a crime. Id. at 21, 88 S.Ct. 1868. In this case, Brogan had substantial information supporting a suspicion that Stephens, whom he then knew only as Judge, had attempted to pass a bad check. Stephens, first, had offered a check that was rejected; second, had done so in a manner that aroused the suspicions of store employees, who called the police; and, third, had run from police officers who approached him after he returned to collect his identification. An individual’s flight from police combined with other observations by a police officer may support reasonable suspicion sufficient for detention under Terry. Illinois v. Wardlow, 528 U.S. 119, 120 S.Ct. 673, 145 L.Ed.2d 570 (2000). In Wardlow itself, the only relevant fact other than flight known to the detaining officer was the suspect’s presence in an area known for crime. Id. at 124-25, 120 S.Ct. 673. Stephens’s prior behavior at the Circuit City store suggested guilt more strongly than would simple presence in such an area, so that the police knew more than was necessary to justify temporarily detaining Stephens. Moreover, Brogan knew enough to suspect reasonably that Scott was involved in the same crime. Brogan knew from Hom-si that Scott was driving the same car that had earlier dropped Stephens off at Circuit City, and he knew from his own observation that Stephens had run towards the car. He had also noted that Scott had parked further from the store than one ordinarily would if merely driving a friend on an innocent errand. These observations, however generously read, do not approach probable cause; but they satisfy, as the district court concluded, the requirement that the police know specific and articulable facts on which they may base a temporary detention. Brogan’s actions prior to his frisk of Scott and sweep of the car were therefore legal. But the 1997 warrants rely on what was found in Scott’s car in 1995, and so the analysis continues. We are less sanguine about the next stage of the encounter: Brogan’s frisk of Scott and protective sweep of the car. A police officer may frisk a suspect— that is, search the suspect’s person for weapons — on reasonable suspicion that the suspect is armed and dangerous. Terry, 392 U.S. at 27, 88 S.Ct. 1868. A search of the passenger compartment of a car in which a suspect rides requires the same degree of suspicion. Michigan v. Long, 463 U.S. 1032, 1049-51, 103 S.Ct. 3469, 77 L.Ed.2d 1201 (1983). When the officer suspects a crime of violence, the same information that will support an investigatory stop will without more support a frisk, Terry, 392 U.S. at 33, 88 S.Ct. 1868 (Harlan, J., concurring), and therefore by implication a car search. This Circuit has extended that rule to encompass crimes commonly associated with violence, even though the criminal act itself may be nonviolent; an example is large-scale trafficking in illegal drugs. United States v. Gilliard, 847 F.2d 21, 25 (1st Cir.1988) (citing United States v. Trullo, 809 F.2d 108, 113 (1st Cir.1987)). The investigation involved no more than the fraud of attempting to pass a bad check. The district court nevertheless concluded that Brogan's frisk of Scott and sweep of the car were permissible, relying on United States v. Edwards, 53 F.3d 616 (3d Cir.1995), for the proposition that reasonable suspicion of fraud justifies a frisk. The Third Circuit reasoned in Edwards that the perpetrators of a fraud in broad daylight may well arm themselves to make an escape. Id. at 618. This logic would seem to permit a frisk on reasonable suspicion of almost any felony, and we do not follow it today. Nor do the other facts relied upon by the district court seem to us to establish reasonable suspicion that Scott was armed and dangerous. The district court could fairly credit Brogan's testimony that the car was parked some distance from the store, that there was little traffic in the parking lot, that Brogan was alone in confronting Scott, and that Scott knew that the police had already detained Stephens. These facts, however, provide an insufficiently particular basis for Brogan to suspect reasonably that Scott was armed and dangerous, and so do not support the district court's conclusion that the frisk and sweep were legal. See Sibron v. New York, 392 U.S. 40, 63-66, 88 S.Ct. 1889, 20 L.Ed.2d 917 (1968) ("[To conduct a] self-protective search for weapons, [an officer] must be able to point to particular facts from which he reasonably inferred that the individual [searched] was armed and dangerous."). b. Inevitable Discovery That, though, does not end the analysis. We also consider an alternative ground of decision presented by the district court. Although evidence derived from unlawful searches is generally subject to suppression, see Wong Sun, 371 U.S. at 484-87, 83 S.Ct. 407, there are numerous exceptions to this rule. One such, the inevitable discovery exception, applies to any case in which the prosecution can show by a preponderance of the evidence that the government would have discovered the challenged evidence even had the constitutional violation to which the defendant objects never occurred. Nix v. Williams, 467 U.S. 431, 440-48, 104 S.Ct. 2501, 81 L.Ed.2d 377 (1984). The district court found that exception to apply to Scott's case, and we agree. This Circuit discussed the rule of inevitable discovery at length in United States v. Silvestri, 787 F.2d 736 (1st Cir.1986), and has subsequently relied upon Silvestri `s discussion in numerous cases, including United States v. Zapata, 18 F.3d 971, 978 (1st Cir.1994), and United States v. Ford, 22 F.3d 374, 377-81 (1st Cir.1994). Silvestri divided the question of inevitable discovery into three parts: first, whether "the legal means [are] truly independent"; second, whether "both the use of the legal means and the discovery by that means [are] truly inevitable"; and, third, whether "the application of the inevitable discovery exception either provide[s] an incentive for police misconduct or significantly weaken[s] fourth amendment protection." 787 F.2d at 744. The district court reasoned that the questioning of Stephens would have led inevitably to Scott's arrest, even had the illegal search of Scott's car never occurred. Stephens's admission of his true age and place of residence allowed the officers to infer his use of false identification, which gave them probable cause to arrest him for attempting to pass a bad check. Scott's lie to Brogan that he did not know Stephens would then have extended probable cause from Stephens to Scott, and Scott would have been arrested and his car subjected to a routine inventory search. The district court's conclusions about what would have occurred are correct. Thus, the alternate avenue of inquiry that involved questioning Stephens was truly independent, because it did not rely on any information derived from Scott; and it was truly inevitable, because the district court concluded without clear error that Brogan would have detained Scott until he finished questioning Stephens, and the results of that questioning are a matter of historical fact. See Nix, 467 U.S. at 444 n. 5, 104 S.Ct. 2501 (emphasizing the importance of “demonstrated historical facts”). The first two questions posed by Silves-tri are thus answered in the affirmative, save for a difficulty we now reach. The Supreme Court and this Circuit have invariably stated the doctrine of inevitable discovery as requiring inevitable discovery by “legal” or “lawful” means. E.g., Nix, 467 U.S. at 444, 104 S.Ct. 2501; Ford, 22 F.3d at 377-79; Silvestri, 787 F.2d at 738-46. None of those cases involved a third party and a defendant’s claim that the inevitable discovery doctrine should not apply because the third party’s rights had been violated. This case involves such a claim because the district court concluded that the police illegally arrested Stephens without probable cause and that Brogan then illegally questioned him without the warnings required by Miranda v. Arizona, 384 U.S. 436, 86 S.Ct. 1602, 16 L.Ed.2d 694 (1966). The government has not argued to us that the district court was wrong, and so we assume that Stephens’s statements, on which the government’s claim of inevitable discovery rests, were taken in violation of Stephens’s Miranda rights. The context of the district court’s conclusion is important. The police did not believe that they had arrested Stephens before he made the statements; they thought the arrest came later, after the statements, and that the Miranda warnings were timely given. Only on later judicial review did a court determine that the arrests came earlier. We have no way to know whether Stephens would have voluntarily provided any identifying information in the course of the Terry stop had the sequence of events been only slightly different. Ordinarily, that Stephens did not receive Miranda warnings would not benefit Scott in his attempts to exclude evidence. Any illegality did not violate Scott’s personal rights, and courts have restricted the exclusionary rule both in the context of searches and in the context of Miranda to violations of a defendant’s personal rights. Rakas v. Illinois, 439 U.S. 128, 140-50, 99 S.Ct. 421, 58 L.Ed.2d 387 (1978) (applying this principle to an illegal search); United States v. Lopez, 709 F.2d 742, 745 n. 3 (1st Cir.1983) (applying it in passing to a failure to give Miranda warnings). Scott’s case, however, presents the distinct question whether the government in showing inevitable discovery may rely on an illegal action that did not violate the relevant defendant’s personal rights. We know of no rule that the “legality” aspect of the inevitable discovery doctrine plays no role in the analysis when third party rights are involved. The deterrence rationale may be significant in such cases as well. We think this question is close. The application of the inevitable discovery exception to this case would allow the government to benefit at least somewhat from the unconstitutional actions of the Natick police — and if here there were two illegalities rather than one, that arguably strengthens rather than weakens the need for suppression as a means of deterrence. The history of the inevitable discovery exception, however, is instructive. The Supreme Court has described this exception as an “extrapolation,” Murray v. United States, 487 U.S. 533, 539, 108 S.Ct. 2529, 101 L.Ed.2d 472 (1988), from the older principle of the independent source. This principle, discussed by Justice Holmes in Silverthone Lumber Co. v. United States, 251 U.S. 385, 40 S.Ct. 182, 64 L.Ed. 319 (1920), is that although the government may first learn of certain facts by illegal means, it may nevertheless prove those facts at trial if it later learns of them by independent, legal means. Id. at 392, 40 S.Ct. 182 (“Of course this does not mean that ... facts [illegally] obtained become sacred and inaccessible. If knowledge of them is gained from an independent source they may be proved like any others.... ”). We think, moreover, that in the case of an actual discovery by an independent source, a defendant cannot obtain the remedy of suppression by simply relying solely on an illegality related to that source if the illegality did not violate the defendant’s personal rights. There is no per se rule operating in either direction. An independent source permits use of the evidence because it breaks the causal chain between the constitutional violation alleged and the discovery of the evidence challenged. See Segura v. United States, 468 U.S. 796, 814, 104 S.Ct. 3380, 82 L.Ed.2d 599 (1984) (“[E]vidence will not be excluded as ‘fruit’ unless the illegality is at least the ‘but for’ cause of the discovery of the evidence.”); see also United States v. Crews, 445 U.S. 463, 471, 100 S.Ct. 1244, 63 L.Ed.2d 537 (1980) (“In the typical ‘fruit of the poisonous tree’ case ... the question before the court is whether the chain of causation proceeding from the unlawful conduct has become so attenuated or has been interrupted by some intervening circumstance so as to remove the ‘taint’ imposed upon that evidence by the original illegality.”). The question in independent source cases is one of causation; suppression requires at least a finding that the challenged evidence would not have been obtained but for a constitutional violation as to the defendant in the case at issue. We conclude, because of the close link between the two doctrines, that the principle should be the same in the case of inevitable discovery: a means by which challenged evidence would inevitably have been discovered that itself violates the law is not, by that violation alone, unlawful as to a defendant if those means did not violate that defendant’s personal rights. Silvestri’s third question, however, focuses the issue: “application of the inevitable discovery exception [must not] provide an incentive for police misconduct or significantly weaken fourth amendment protection.” 787 F.2d at 744. Such an analysis necessarily dwells closely on the facts of a particular case. Moreover, a court in conducting this analysis must bear always in mind the social costs of the exclusionary rule. Cf. People v. Defore, 242 N.Y. 13, 150 N.E. 585, 587 (1926) (Cardozo, J.) (“The criminal is to go free because the constable has blundered.”). We look at the incentives here. In this case, there were two unconstitutional acts, rather than one, and we think it appropriate to take Brogan’s treatment of Stephens into account in assessing the facts as a whole with regard to Scott. On these facts, Brogan had little incentive to violate Stephens’s Miranda rights (had he known he was doing so) in order to gain the advantage with Scott of use of the inevitable discovery doctrine. Brogan’s incentive was to be able to use Stephens’s statements against him. Application of the inevitable discovery doctrine would not, we think, act here as an incentive to unconstitutional behavior. Other cases may present different incentives and warrant a different outcome. We also take into account that neither constitutional violation was truly egregious, as was the excessive force used in United States v. Rullo, 748 F.Supp. 36 (D.Mass.1990). Moreover, although we have assumed the correctness of the district court’s view that the police subjected Stephens to a de facto arrest, we think that point was itself close and that the police might have reasonably believed they had undertaken only a justified investigatory stop. We hold that on the facts of this case the doctrine of inevitable discovery applies and exclusion is not required to remove incentives for future police misconduct. 2. The 1997 Search Pursuant to Warrant The district court found that even were the results of the 1995 search subject to exclusion, the remainder of the 1997 warrant provided probable cause and the information in it would inevitably have been discovered through independent means. We need not reach that argument given our conclusion that the 1995 search should not have been suppressed. The information from 1995 and thereafter provided sufficient probable cause. Scott makes an independent argument that the affidavit was nonetheless materially misleading. We agree with the district court’s reasoning that none of the claimed errors were material, none of the claimed omissions contradicted the sworn facts, and neither the errors nor the omissions affected the probable cause determination. The district court was therefore also correct to conclude that Scott had not made a sufficient showing to warrant a hearing under Franks v. Delaware, 438 U.S. 154, 98 S.Ct. 2074, 57 L.Ed.2d 667 (1978), which established the right to such a hearing for a defendant who shows some evidence of intentional or reckless material misrepresentation in the affidavit supporting a search warrant. See id. at 171-72, 98 S.Ct. 2674 (“[I]f, when material that is the subject of the alleged falsity or reckless disregard is set to one side, there remains sufficient content in the warrant affidavit to support a finding of probable cause, no hearing is required.”). Scott also argues that there was no probable cause to search his car, a Mercury Mountaineer, even if there was cause to search his house. As did the magistrate and the district court, we disagree. The warrant affiant had seen a three-inch-thick stack of documents in the back seat, hidden by a newspaper, and had seen other codefendants load things into the car. There was ample reason- — especially in light of the 1995 search, which had revealed incriminating documents in the trunk of another of Scott’s cars — to think the car was used to transport his runners and, likely, the documents essential to his trade. That is particularly so when the initial reasoning was that of an experienced field agent who had watched the growing sophistication of Scott’s schemes over the years. The magistrate who issued the warrant was entitled to consider, the agent’s expertise. United States v. Zayas-Diaz, 95 F.3d 105, 111, 116 (1st Cir.1996). B. Evidentiary Rulings We review a district court’s decision to admit or to exclude evidence for abuse of discretion. United States v. Gilbert, 181 F.3d 152, 160 (1st Cir.1999). Even if we find error, we will not reverse if the error was harmless — that is, if it is highly probable that the error did not contribute to the verdict. United States v. Tse, 135 F.3d 200, 209-10 (1st Cir.1998). If, moreover, the defendant failed at trial to object to introduction of the evidence, we review only for plain error. United States v. Moore, 923 F.2d 910, 915-16 (1st Cir. 1991). I. Other Crimes The jury heard evidence of other crimes or possible crimes Scott committed on four occasions. Litigants may not use such evidence to prove that a person was of bad character and acted in conformity with that character. Fed.R.Evid. 404(b). They may, however, use the same evidence to prove other facts at issue in the case, including “motive, opportunity, intent, preparation, plan, knowledge, identity, or absence of mistake or accident.” Id. This list illustrates, but does not exhaust, permissible purposes for such evidence. E.g., Udemba v. Nicoli, 237 F.3d 8, 15 (1st Cir.2001) (permitting the use of such evidence to show "the extent of damages attributable to emotional distress"). On none of the four occasions did the district court err. First, Brogan testified that he arrested Scott and Stephens in 1995 and described what he found in Scott's car. As Scott was accused of conspiring with Stephens, the evidence was relevant to show "the background, formation, and development of the illegal relationship" between the two. United States v. Varoudakis, 233 F.3d 113, 118 (1st Cir.2000) (quoting United States v. Escobar-de Jesus, 187 F.3d 148, 169 (1st Cir.1999)). Furthermore, the district court gave a cautionary instruction that the evidence was meant to put other evidence in context, directing the jury not to speculate as to the reasons for the arrest and telling it that the arrest had "nothing to do with this indictment." There was no error. Second, Charles Lynch, the facility manager at Coolidge House, a federal halfway house in Boston, testified that Scott had been a resident there, which meant that Scott had previously been an inmate in the federal prison system. Coolidge House was close to several ATM machines, and those machines were used to withdraw money from the USAA Federal Savings Bank account used in the scheme-in one instance, only a few minutes after Scott had signed out of the house. The evidence was relevant and admissible under Rule 404(b) because it showed Scott's opportunity to make withdrawals from the bank account, which linked him to the fraudulent tax returns. Also, the district court mentioned this cvi-dence in its cautionary instruction. There was no error. Third, the government introduced over objection two tax returns in the names of Robert Maguire and David Schmidt, which were not charged in the indictment. Other evidence tied those two returns to the returns which were the subject of the indictment. The filing fees for the Maguire and Schmidt returns were paid using the same credit card used to pay the fee of one indicted false return; furthermore, Scott tried to eliminate those charges from the credit card account. The evidence tended to show Scott was in control of the scheme. There was no error in its admission. Fourth, Sonia Subia, a records custodian from the USAA Bank, testified as to various deposits into the Barnes USAA account. Subia's testimony included a reference to four deposits of Massachusetts tax refunds. Scott did not then object; the following morning, he objected that the reference was erroneously admitted because the government had not listed those accounts in its 404(b) letter. The government then explained that it did not mean to argue that those returns were necessar-fly fraudulent; rather, regardless of any fraud, their mere existence showed Scott's control of the scheme and the instrumen-talities of the scheme. The district court found the matter too peripheral to warrant a special jury instruction. Although a special instruction might have been proper had Scott's initial objection been timely, the court neither plainly erred by failing to give one on its own motion nor abused its discretion by refusing one the next day. There was no error. 2. Opinion Testimony Identifying Handwriting The final evidentiary challenge is more serious. James Donahue, an IRS agent, testified that certain documents in evidence were in Scott’s handwriting. Donahue had followed Scott through a number of IRS investigations over a number of years, beginning when Donahue participated in the 1989 search. Donahue testified on voir dire that during this time he had seen examples of Scott’s handwriting that included three to five letters, five to ten court pleadings, signature cards to open three bank accounts in Scott’s name, fifty or sixty checks and deposit slips for those accounts, five to ten money orders, applications to file tax returns electronically, two driver’s licenses, a pilot’s license, and five to ten forms Scott had signed as part of booking procedures at police stations, among others. Donahue also saw Scott sign a document in his presence: a fingerprint card at the marshal’s office during an earlier investigation. Scott made two objections to Donahue’s testimony as a whole, which he renewed after the voir dire. First, he objected that Donahue had acquired familiarity with Scott’s handwriting “for purposes of the litigation,” and thus could not testify as a lay witness under Federal Rule of Evidence 901(b)(2). Second, he objected that despite Donahue’s exposure to Scott’s handwriting Donahue nevertheless lacked sufficient familiarity with that handwriting to testify. The district court overruled Scott’s objections, but required Donahue to refer to the documents in general terms in order to avoid undue prejudice to Scott. The jury was not told that the documents providing the basis for Donahue’s familiarity came from earlier investigations, arrests, and seizures. Donahue then identified Scott’s signature and handwriting on various documents which the government had previously introduced as found in Scott’s possession. These documents included a birth certificate in the name of Ralph Swoboda, a list of names, which included Scott’s codefendants, the purported signature of Randy LaPlante on Western Union forms, an application for a copy of Daniel Richard Brogan’s birth certificate, and similar papers. To some of these Scott objected on an individual basis as unduly prejudicial under Federal Rule of Evidence 403; to others, he did not. The district court overruled all such objections. In admitting the testimony, the district court relied on Federal Rule of Evidence 701, which reads: If the witness is not testifying as an expert, the witness’ testimony in the form of opinions or inferences is limited to those opinions or inferences which are (a) rationally based on the perception of the witness, and (b) helpful to a clear understanding of the witness’ testimony or the determination of a fact in issue, and (c) not based on scientific, technical, or other specialized knowledge within the scope of Rule 702. Fed.R.Evid. 701. The district court distinguished Rule 901 as dealing only with authentication as a condition precedent to the admissibility of evidence, and therefore not with evidence that had already been admitted. The court reasoned in the alternative that if Rule 901 did apply, Rule 901(b)(2) (on which Scott relied) would nevertheless permit admission. The Rule states: By way of illustration only, and not by way of limitation, the following are ex- . ampies of authentication or identification conforming with the requirements of this rule: (2) Nonexpert opinion on handwriting. Nonexpert opinion as to the genuineness of handwriting, based upon familiarity not acquired for purposes of the litigation. Fed.R.Evid. 901(b). Scott argues primarily that the correct rule for considering admissibility of such handwriting authentication or identification evidence is Rule 901(b)(2), and that the evidence was inadmissible because Donahue came to know Scott’s handwriting through his criminal investigation of Scott. Rule 901(b) illustrates two ways of authenticating or identifying through the testimony of a witness the handwriting on a document as being written by a particular person. The handwriting may be identified through a lay witness who has familiarity with the alleged author’s handwriting and who did not acquire that familiarity for purposes of the litigation. Fed.R.Evid. 901(b)(2); United States v. Tipton, 964 F.2d 650, 655 (7th Cir.1992) (surveying cases). Alternatively, the handwriting may be identified as the alleged author’s handwriting by an expert, who lacks such familiarity except as acquired for purposes of the litigation but has the requisite expertise and who compares the sample with specimens which have been authenticated. Fed.R.Evid. 901(b)(3); see also Fed.R.Evid. 901(b)(2) advisory committee’s note (“Testimony based upon familiarity acquired for purposes of the litigation is reserved to the expert....”). The district court ruled that Rule 701 governed and that Rule 901 did not apply. This is an issue of law we review de novo. We hold that both rules must be satisfied. The essence of the district court’s reasoning was that the documents that contained the handwriting had already been admitted into evidence. As such, they had, of course, been authenticated or identified. Fed.R.Evid. 901(a). Authentication refers to evidence that tends to prove that a document is what its proponent claims it to be. The documents were earlier authenticated as documents found in the search of Scott’s car or home, and admitted as such. The handwriting on those documents was not previously authenticated as evidence that tended to prove that the handwriting was Scott’s. To prove that point, the evidence had to comply with Rule 901. Whether Donahue’s testimony satisfied Rule 901 depends on the purpose of the requirement that the familiarity of a lay opinion witness with handwriting may not be “acquired for purposes of the litigation.” This limitation is properly understood in light of the common law tradition with which Rule 901(b)(2) and Rule 901(b)(3) break. English courts, and early American courts, placed strict limits on testimony concerning identification of handwriting by its style, called “comparison of hands.” 7 J. Wigmore, Evidence § 1991, at 253 (Chadbourn rev.1978). Professor Wigmore explains that when this rule was at its strictest, no witness was permitted to identify handwriting in a criminal case unless the witness had seen the document in question written or signed. Over time, however, courts came to permit handwriting identification based on ever-looser degrees of familiarity: one who had seen the alleged author sign other documents, or one who had seen samples of writing known to be by that author over a period of time, could give such testimony. Id. §§ 1992-1993, at 257-62. Rule 901(b)(2) retains only one vestige of the common law rule. A lay witness may not enter court, see for the first time two samples of handwriting, and identify the contested sample as written by the same person as the previously authenticated sample; the result is the same, moreover, if the witness compared the two samples before entering the courtroom. See, e.g., United States v. Pitts, 569 F.2d 343, 348 (5th Cir.1978) (affirming the exclusion of the testimony of a lawyer-witness who had compared a single contested sample of handwriting to a single authenticated one in preparing for a previous trial). After all, such a comparison could be made as easily by the jury as by the witness. Therefore, lay witness testimony without familiarity would not be helpful to the jury and would be prohibited by Rule 701 even if Rule 901 did not exist. Cf. Wigmore, supra, § 1993, at 260 (observing that the common-law opinion rule would preclude lay witness testimony without familiarity). This case involves no such situation. Donahue became familiar with Scott’s handwriting over the course of several years, and he did so not for the purpose of testifying, but instead for the purpose of solving a crime. Scott was perfectly entitled to argue to the jury that Donahue’s interest in securing a conviction colored Donahue’s perception of Scott’s handwriting. Cf. Johnson v. United States, 333 U.S. 10, 14, 68 S.Ct. 367, 92 L.Ed. 436 (1948) (discussing in another context the pressures facing those “engaged in the often competitive enterprise of ferreting out crime”). That possibility, however, did not require the district court to exclude the evidence under Rule 901(b)(2). We may dispense quickly with Scott’s remaining objections to Donahue’s testimony. Scott makes a different argument that Donahue was not sufficiently familiar with Scott’s signature to testify. This fact-bound ruling is reviewed for abuse of discretion. Gilbert, 181 F.3d at 160. It is true that Donahue conceded that he saw Scott sign his name only once. Other categories of experience can, however, demonstrate familiarity, such as seeing signatures on writings purporting to be those of the alleged author when the circumstances would indicate that they were genuine. 2 McCormick on Evidence § 221, at 42 (J. Strong ed., 5th ed.1999). The rule for which Scott seems to contend — not only a strict requirement that the witness see the alleged author in the act of signature, but also a further demand that this occur multiple times — is far too strict. See United States v. Standing Soldier, 538 F.2d 196, 202 (8th Cir.1976) (permitting authentication testimony on the basis of a single exposure to an uncontested signature and a single exposure to a contested one). In this case, there was enough familiarity on the part of the witness to admit the testimony. Scott’s arguments go rather to the weight of the evidence. Weinstein’s Federal Evidence § 901.04[2], at 901-20 (J. McLaughlin ed., 2d ed.2001). Scott took advantage of this, and cross-examined the agent on the paucity of actual sightings of Scott endorsing his signature. Accordingly, there was no abuse of discretion and no error. Scott also argues he was prejudiced by the implicit message, conveyed by the use of an IRS agent to identify his handwriting, that the government had been investigating him for a long time and probably for other things. The district court admirably restricted direct testimony to this effect. Nevertheless, Scott would have us say that the remaining prejudice still substantially outweighed the probative value of the evidence so that the district court was bound to exclude the testimony under Rule 403. We cannot so hold. The district court’s discretion in the Rule 403 balancing inquiry is broad. E.g., Daigle v. Me. Med. Ctr., Inc., 14 F.3d 684, 690 (1st Cir.1994). Even if the government might have done better to use an expert witness for the handwriting identification, as Scott argues, and even if another district court might permissibly have excluded the evidence on this basis, the availability of a less prejudicial method of proof is only a factor to be weighed in the Rule 403 inquiry and does not control this case. See Fed.R.Evid. 403 advisory committee’s note (“The availability of other means of proof may also be an appropriate factor.”) (emphasis added); see generally Old Chief v. United States, 519 U.S. 172, 180-92, 117 S.Ct. 644, 136 L.Ed.2d 574 (1997) (discussing the proper analysis of less prejudicial methods of proof). In light of its careful attention to the problem of prejudice, the district court did not abuse its discretion by admitting the evidence under Rule 403. C. Sentencing Issues 1. Leadership Role in the Offense The district court’s conclusion that Scott played a leadership role and its enhancement of his sentence under USSG § 3B1.1 rest largely on the facts of the case and we therefore review for clear error. United States v. Graciani, 61 F.3d 70, 75 (1st Cir.1995); see also United States v. Wright, 873 F.2d 437, 443-44 (1st Cir.1989) (discussing the rationale for reviewing §§ 3B1.1 and 3B1.2 determinations for clear error). A