United States v. Khan Mohammed

United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT Argued November 17, 2011 Decided September 4, 2012 Reissued September 21, 2012 No. 09-3001 UNITED STATES OF AMERICA, APPELLEE v. KHAN MOHAMMED, APPELLANT Appeal from the United States District Court for the District of Columbia (No. 1:06-cr-00357-1) Shardul S. Desai argued the cause for appellant. With him on the briefs was Peter S. Spivack, appointed by the court. Vijay Shanker, Attorney, U.S. Department of Justice, argued the cause for appellee. With him on the brief were Lanny A. Breuer, Assistant Attorney General, and Matthew Robert Stiglitz, Attorney, U.S. Department of Justice. Kevin R. Gingras and Teresa A. Wallbaum, Attorneys, U.S. Department of Justice, and Roy W. McLeese III, Assistant U.S. Attorney, entered appearances. [2] Before: SENTELLE, Chief Judge, GRIFFITH and KAVANAUGH, Circuit Judges. Opinion for the Court filed by Circuit Judge GRIFFITH. Opinion concurring in part and concurring in the judgment filed by Circuit Judge KAVANAUGH. GRIFFITH, Circuit Judge: Khan Mohammed challenges his conviction and life sentence for narcoterrorism. He also claims that his trial counsel provided ineffective assistance. We affirm Mohammed’s conviction and sentence but remand for the district court to hold an evidentiary hearing on the claim of ineffective assistance. I While living in Pakistan in 2006, a man named Jaweed, who hailed from the village of Geratak in Afghanistan’s Nangarhar province, fell in with Abdul Rahman, a former Taliban official for the Jalalabad province of Afghanistan also living in Pakistan. Rahman was plotting an attack on the Jalalabad airfield, a strategic NATO airbase in eastern Afghanistan, and instructed Jaweed to return to Geratak and contact a fellow villager, Khan Mohammed, who was also involved in the plot and needed help. Jaweed did as he was told and visited Mohammed, who brought him into the planning of the attack, directing him to obtain the missiles that would be used in the strike. But Jaweed soon turned against Rahman and Mohammed and disclosed the plot to Afghan authorities. The Afghan police persuaded Jaweed to continue his role in the plot, but to become their informant. When primary responsibility for the [3] investigation was turned over to agents of the U.S. Drug Enforcement Administration (DEA) deployed in Nangarhar, Jaweed worked with them as well. The DEA agents wired Jaweed and recorded several of his conversations with Mohammed in August and September 2006. In the first of those conversations, Mohammed discussed with Jaweed details of the attack on the airfield and claimed that he had not only the same purpose as Rahman, but the same authority. Hearing his plans and his boast, the DEA decided to arrest Mohammed soon after Jaweed had given him the missiles. Concern about losing control of the missiles once they were in Mohammed’s hands, however, led to a different strategy. The DEA would arrest Mohammed for narcotics trafficking. Following this plan, Jaweed told Mohammed he had a friend looking for opium. Mohammed replied that he knew a source who could supply as much as Jaweed’s friend needed. Jaweed and Mohammed met three more times to iron out details, such as the price for the opium and Mohammed’s commission. These discussions also included plans for the attack on the airfield. For example, during one of the meetings Mohammed said they needed a car to secure the missiles. See Government Trial Ex. 2C (Mohammed, stating that they would “tightly and firmly load [the missiles] in our car”). Eleven days later, Mohammed announced at another meeting that he would use the profits from the drug sale to buy a car, which could help carry out more drug deals. The opium deal went off without a hitch. Jaweed accompanied Mohammed to a local bazaar where Mohammed negotiated the sale with his source. The next day, Jaweed accompanied Mohammed to the seller’s home and secretly videotaped Mohammed inspecting, paying for, and taking away the opium he then sold to Jaweed. Pleased with the results, the [4] DEA agents told Jaweed to orchestrate another sale, this time for heroin. When Jaweed raised the idea, Mohammed readily agreed and acquired almost two kilograms of heroin, which he then sold to Jaweed. Mohammed was enthused by the prospect of how much money their newly formed drug business could make. When Jaweed told Mohammed that his friend would send the opium and heroin to the United States, Mohammed declared, “Good, may God turn all the infidels to dead-corpses.” Government Trial Ex. 2H. Their “common goal,” Mohammed told Jaweed, was to eliminate the “infidels” either “by opium or by shooting.” Id. On October 29, 2006, the DEA and Afghan police arrested Mohammed at a roadside checkpoint. They blindfolded and handcuffed him and drove him to a DEA base at the Jalalabad airfield. He was briefly held in a detention cell without handcuffs or blindfold and then taken to a room to be questioned about the drug deals and the planned attack on the airfield. During his interrogation, Mohammed was neither blindfolded nor shackled. The record is unclear whether he was handcuffed. Three DEA agents conducted the questioning, one wearing a visible sidearm. Speaking through an interpreter, DEA Special Agent Jeffrey Higgins read Mohammed the Miranda warnings, which were translated into Pashto, his native language. Higgins asked Mohammed if he understood the warnings. Mohammed said that he did. Thinking Mohammed illiterate, Higgins did not ask him to sign a rights card or a waiver form. Mohammed told Higgins that he was willing to answer questions, and the interview began. It lasted about two hours, including a short break. Mohammed showed no distress during the questioning or any difficulty understanding the interpreter. He was given food and water, a prayer rug and the opportunity to pray, and was allowed [5] to use the bathroom. Questioning took place in a conversational tone and none of the DEA agents threatened Mohammed, although one told him, falsely, that his hands had tested positive for heroin. At no time did Mohammed ask for an attorney or that the questioning stop. At the end of the interrogation, the DEA agents told Mohammed that he was being charged with drug trafficking and terrorism under U.S. law. He was later transferred to the United States for trial. With trial scheduled for May, in January 2008 Mohammed moved to suppress the statements he made during his questioning at the airfield on the ground that his Miranda waiver was involuntary. On April 30, the district court denied his motion, concluding that he understood the Miranda rights and the consequences of giving them up. Tr. Status & Mots. Hr’g 18:12-18, Apr. 30, 2008. On May 1, with jury selection set to begin in a week, Mohammed moved for a three-month continuance of the trial. Summoned to an ex parte hearing the next morning, Mohammed’s counsel expressed to the court his fear that he had “been ineffective in assisting [Mohammed]” by failing to follow up on his lead that witnesses in Afghanistan would rebut the government’s allegation that he was part of the Taliban. Tr. Ex Parte Sealed Discussions 5:23-6:3, May 2, 2008. Counsel also asked for more time to prepare to examine the government’s narcoterrorism expert. Hearing this, Mohammed asked the court for new counsel. The court convened a second ex parte hearing that afternoon to address his request. Speaking directly to the court at that hearing, Mohammed argued that his attorney could have obtained evidence that he was not part of the Taliban and that Jaweed was a thief. The court reminded Mohammed that he had already lodged the theft allegation against Jaweed, that the court had previously decided it lacked corroboration, and that Mohammed had agreed not to pursue the [6] allegation at trial.1 Mohammed’s counsel added that these witnesses might undermine Jaweed’s credibility more generally by saying he was a liar. But under questioning by the court, he could point to no basis to believe that they would. Asked why he had not already tracked down these potential witnesses, Mohammed’s counsel claimed there were “insurmountable” logistical problems with traveling to Afghanistan to locate and bring them to the United States. Id. at 36:13-37:5. His trial preparation had focused on other issues instead. The court then called yet another hearing that same day, this time with all the parties. Learning of Mohammed’s concerns, the government announced to the court that it would not seek to link him to the Taliban. Hearing that, Mohammed, through his counsel, agreed to withdraw his motion for a continuance and proceed to trial. The government put on its case in four days. Two of these days were taken up with Jaweed’s testimony. Mohammed’s counsel called no witnesses and offered no evidence, and Mohammed did not take the stand. On May 15, 2008, a jury found Mohammed guilty of international drug trafficking, 21 U.S.C. §§ 959 (a)(1), (2), and drug trafficking with intent to provide financial support to a terrorist, id. § 960a. At sentencing, Mohammed objected to the recommendation in the Presentencing Report that the court apply the terrorism enhancement of the Sentencing Guidelines. The court found no [1] While in custody in Afghanistan, Mohammed told his U.S. captors that Jaweed was a convicted thief set to serve an eighteen-year sentence in Afghanistan. The government filed a motion in limine to exclude any such impeachment evidence, arguing that Jaweed had denied this uncorroborated allegation. At Mohammed’s request, the government conducted a background check that turned up no criminal history for Jaweed in the Nangarhar province. Mohammed later agreed to drop this line of argument. [7] basis for the objection and applied the enhancement, but explained that it could have exercised its discretion under § 960a to impose the same sentence even without the enhancement. The district court sentenced Mohammed to two concurrent life sentences. Mohammed does not challenge his conviction for international drug trafficking. He appeals only his conviction and sentencing for narcoterrorism. He also raises a claim of ineffective assistance of counsel. We take jurisdiction under 28 U.S.C. § 1291 and 18 U.S.C. § 3742 . II We consider first Mohammed’s argument that his Miranda waiver was invalid and that the district court erred by denying his motion to suppress the statements he made during his interrogation at the Jalalabad airfield. We need not resolve the novel question whether Miranda applies to the overseas custodial interrogation of a person who is not a U.S. citizen. Even if we assume it does, any alleged error by the district court was harmless because the government made no effort to use the statements at trial. See, e.g., United States v. Patane, 542 U.S. 630, 641 (2004) (plurality opinion) (“Potential [Miranda] violations occur, if at all, only upon the admission of unwarned statements into evidence at trial.”); Oregon v. Elstad, 470 U.S. 298, 306-07 (1985) (noting that the Fifth Amendment prohibits using compelled statements in the prosecution’s case-in-chief). On appeal, Mohammed maintains that his statements were used against him because Higgins was only able to identify Mohammed’s voice on the recordings at trial from having heard it first during the interrogation. But voice identification is not the type of incriminating information Miranda protects: [8] “Requiring a suspect to reveal the physical manner in which he articulates words, like requiring him to reveal the physical properties of the sound produced by his voice, does not, without more, compel him to provide a ‘testimonial’ response for purposes of the [Fifth Amendment] privilege.” Pennsylvania v. Muniz, 496 U.S. 582, 592 (1990) (citation omitted); see also Elstad, 470 U.S. at 317 (noting that Miranda ensures a suspect’s unwarned answers may be excluded from the government’s case-in-chief). Mohammed also asserts that what he said during interrogation was used against him indirectly. The government’s ability to rely on his statements for impeachment purposes, so his argument goes, made him hesitant to testify in his own defense. But this argument has no constitutional weight. Statements taken in violation of Miranda are admissible as impeachment evidence unless they are, in very fact, involuntary. See, e.g., Elstad, 470 U.S. at 307 ; Oregon v. Hass, 420 U.S. 714, 723 (1975) (holding that unwarned statements are admissible for impeachment purposes unless an “officer’s conduct amount[ed] to an abuse,” in which case admissibility is governed by “the traditional standards for evaluating voluntariness and trustworthiness”). And whatever one might conclude about the merits of Mohammed’s Miranda claim, he certainly has not shown the egregious facts necessary to establish that the statements he made during questioning were involuntary. See, e.g., Berghuis v. Thompkins, 130 S. Ct. 2250, 2263 (2010) (finding no coercion in a three-hour interrogation — longer than Mohammed’s — without evidence “that police threatened or injured [the defendant] during the interrogation or that he was in any way fearful”); Mincey v. Arizona, 437 U.S. 385, 398-99 (1978) (finding a confession involuntary when the defendant had been shot and paralyzed a few hours before questioning, was in intense pain, and gave confused and incoherent responses). [9] We have been given no reason to disturb the district court’s findings that the DEA agents did not threaten or intimidate Mohammed, that he was treated well during his relatively brief interrogation, and that he seemed eager to talk and comfortable enough to ask questions when he needed clarification. See Tr. Status & Mots. Hr’g 10:10-12:22. Mohammed emphasizes that he was blindfolded and handcuffed at times and perhaps even handcuffed during questioning. But no court has found that waivers made while a suspect is handcuffed are invalid for that reason alone, see, e.g., United States v. Adams, 583 F.3d 457, 467-68 (6th Cir. 2009) (upholding an implicit Miranda waiver even though the defendant was handcuffed while he was read his rights and during questioning); United States v. Doe, 149 F.3d 634, 639 (7th Cir. 1998) (finding a Miranda waiver voluntary despite questioning the defendant in a remote location while handcuffed in the back of a squad car, with some officers wearing masks), much less that statements obtained while handcuffed are themselves involuntary. And although an agent lied to Mohammed that his hands tested positive for heroin, misleading a suspect during interrogation is only one factor in the totality of the circumstances analysis that governs our inquiry into voluntariness. See Frazier v. Cupp, 394 U.S. 731, 739 (1969). Just as telling a defendant, falsely, that his codefendant had already confessed to murder is “relevant,” yet not enough to render an “otherwise voluntary confession inadmissible,” id., the lie here is insufficient to outweigh the rest of the evidence showing that Mohammed’s statements were voluntary. Within the full context of the interrogation, we would be hard pressed to conclude that the possibility Mohammed was handcuffed combined with the agent’s lie was enough to render his statements involuntary. [10] III Mohammed next argues that the evidence at trial cannot sustain his conviction under 21 U.S.C. § 960a. This statute criminalizes conduct abroad that would violate domestic drug laws if “committed within the jurisdiction of the United States” when the actor “know[s] or intend[s] to provide, directly or indirectly, anything of pecuniary value to any person or organization that has engaged or engages in terrorist activity . . . or terrorism.” Id. Mohammed does not dispute that the evidence was sufficient to prove he engaged in a qualifying drug offense, that he met the statutory definition of a person who engages in terrorism, or that he knew the transaction would result in financial gain to himself. Instead, he urges us to graft an additional, unwritten intent requirement onto the statutory text, which he calls a “drug-terror nexus.” Appellant’s Br. 46. Under this theory, it is not enough that Mohammed committed a drug offense with intent to provide pecuniary value to a terrorist or terrorist organization; the government must also show he knew that the money would support terrorist acts. But Mohammed overlooks the straightforward terms of the statute. Section 960a requires proof that the defendant intended to support a “person or organization that has engaged or engages in terrorist activity,” not that he intended his funds to be used for any particular activity. 21 U.S.C. § 960a (emphasis added). The first step in statutory interpretation considers the statute’s plain language, see United States v. Villaneuva-Sotelo, 515 F.3d 1234, 1237 (D.C. Cir. 2008), and we decline Mohammed’s invitation to ignore the words Congress chose. The text is abundantly clear that Congress intended to target drug offenses the defendant knows will support a “person or organization” engaged in terrorism, with no additional requirement that the defendant intend his drug trafficking to [11] advance specific terrorist activity. In other words, Mohammed need not have planned for his drug proceeds to fund terrorist ends. It is sufficient that the proceeds went to a terrorist — him. Mohammed argues that we must look past this plain language because only his proposed intent requirement saves § 960a from merely duplicating the work of statutes that already criminalize drug trafficking overseas, see 21 U.S.C. § 959 , and material support of terrorism, see 18 U.S.C. §§ 2332d, 2339A, 2339B, 2339C. But the premise that § 960a is redundant is suspect. Congress could have reasonably determined that international drug trafficking combined with the intent to support a terrorist is a different crime — more blameworthy, more dangerous, or both — than drug trafficking overseas and material support of terrorism committed separately. Or Congress could have decided that the ability to charge one crime instead of two was a valuable, perhaps necessary, tool for prosecutors that warranted creating a new crime. In any event, Congress need not act with the sort of precision Mohammed’s argument assumes: “Redundancies across statutes are not unusual events in drafting,” and courts must give effect to overlapping statutes unless there is “positive repugnancy” between them. Conn. Nat’l Bank v. Germain, 503 U.S. 249, 253 (1992) (quoting Wood v. United States, 41 U.S. (16 Pet.) 342, 363 (1842)) (internal quotation marks omitted). Mohammed emphasizes that § 960a’s penalty is greater than those for drug trafficking and material support combined, but that alone does not establish “positive repugnancy.” See Wood, 41 U.S. (16 Pet.) at 363 (defining “manifest and total repugnancy” as more than “merely affirmative, or cumulative or auxiliary” provisions, but divergence between statutes so strong as “to lead to the conclusion that the latter laws abrogated, and were designed to abrogate the former”); see also United States v. Batchelder, 442 U.S. 114, 123 (1979) (“So long as overlapping criminal [12] provisions clearly define the conduct prohibited and the punishment authorized, the notice requirements of the Due Process Clause are satisfied.”). At most, Mohammed highlights some congressional overlap among statutes directed at international drug trafficking and support of terrorism. That is no reason for us to depart from the clear text of a statute. But absurd results will follow unless we do, Mohammed argues. As an example, he offers the hypothetical of a father who sells drugs to pay a ransom to the Revolutionary Armed Forces of Columbia for his kidnapped child. Mohammed contends that this father, whose paternal love and not any support of terrorism drove him to crime, risks life imprisonment under the government’s reading of § 960a. Appellant’s Br. 46-47. But finding § 960a absurd based on this possibility would have broad implications for criminal law writ large. We can imagine similar problems for any sympathetic defendant forced by his circumstances to break the law. The criminal justice system deals with such unusual fact patterns through prosecutorial discretion and traditional defenses such as the duress defense, but not by rewriting criminal statutes that are uncontroversial in the overwhelming majority of their applications. Similarly, Mohammed argues that, limited to its text, § 960a could reach an individual who donates some portion of his drug proceeds to a person or organization that engaged in terrorist acts in the past, but no longer does so.2 Appellant’s Reply Br. 21-22. Although the statute’s use of the past and present tense — “has engaged or engages in . . . terrorism” — increases [2] This concern does not arise in Mohammed’s case, where the government did not introduce evidence of his past terrorist involvement at trial, but instead relied on evidence that he was planning a terrorist attack. [13] its potential breadth, such a result is by no means absurd. It neither defies “rationality” nor “common sense.” Landstar Express America, Inc. v. Fed. Mar. Comm’n, 569 F.3d 493, 498 (D.C. Cir. 2009); Suburban Transit Corp. v. I.C.C., 784 F.2d 1129, 1130 (D.C. Cir. 1986). Wide-reaching criminal statutes are common, and while reasonable minds may differ about the wisdom of § 960a’s scope, “debatable policy . . . is hardly irrational,” Landstar, 569 F.3d at 499 ; cf. United States v. Ramsey, 165 F.3d 980, 990 (D.C. Cir. 1999) (rejecting a proposed interpretation for its “obvious absurdities” of ending “a centuries-old practice” in the criminal justice system and exposing federal prosecutors and judges to liability for entering into and approving plea agreements). Moreover, even if we were persuaded that the statute is redundant or leads to absurd results that justify a departure from its plain meaning, Mohammed’s proposed solution is utterly without support. The text lends no aid, as we have already discussed, and even his resort to legislative history is unavailing. He leans on three statements from some of the statute’s supporters to argue that some members of Congress believed § 960a’s purpose is to punish those who use proceeds from drug sales to support terrorism. Appellant’s Br. 47-48 (quoting Rep. Hyde’s statement that Congress intended § 960a to “address and punish those who would use . . . illegal narcotics to promote and support terrorism,” 151 CONG. REC. H6292 (daily ed. July 21, 2005), and statements from Reps. Hyde and Souder that § 960a would address the overlapping links between drug trafficking and global terrorism, id.; id. at H6293). Putting to one side the usual concerns about using legislative history, especially to avoid the plain meaning of a statute, these statements do not even contradict what the statute says. It is clear that Congress intended to punish those who support terrorism directly, as the Congressmen said, as well as indirectly, as the statute provides. [14] IV If we sustain his conviction, Mohammed argues that we should remand for resentencing because the district court erred by applying the terrorism enhancement in the Sentencing Guidelines to calculate his sentencing range. We disagree. The terrorism enhancement, found in Guidelines § 3A1.4(a), increases by twelve the base offense level for calculating a sentencing range if the defendant was convicted of a crime that “involved, or was intended to promote, a federal crime of terrorism.” “Federal crime of terrorism” is defined in 18 U.S.C. § 2332b(g)(5) as an offense in violation of certain enumerated statutes that is “calculated to influence or affect the conduct of government by intimidation or coercion, or to retaliate against government conduct.” Mohammed concedes that § 960a is among the enumerated statutes, but argues that fact alone does not make his offense a “federal crime of terrorism.” In his view, only the mens rea requirement he has urged us to read into the statute — an intent to finance terrorism — would justify including § 960a as a “federal crime of terrorism.” And as Mohammed points out again, the jury was not asked whether he had that intent. But 18 U.S.C. § 2332b(g)(5) offers no support for Mohammed’s theory. The definition of “federal crime of terrorism” contains its own intent element, with an additional requirement only that the offense of conviction appear on the statutory list, as § 960a does. The only question remaining, then, is whether we can sustain the district court’s finding that Mohammed had the requisite intent under § 2332b(g)(5). The district court found two alternate bases to conclude that he did: he “specifically intend[ed] to use the commission from the drug sales to purchase a car to facilitate attacks against U.S. and [15] foreign forces in Afghanistan,” and he “specifically intend[ed] and [was] motivated by the drugs’ destructive powers on U.S. civilian populations as a means of violent jihad against Americans who have fighting forces in Afghanistan against the Taliban.” Sentencing Tr. 17:8-15, Dec. 22, 2008. We conclude that the first finding was sufficient to apply the terrorism enhancement. Mohammed maintains that the evidence does not establish that he intended to use the drug proceeds to buy a car to aid in the Jalalabad attack. In one meeting with Jaweed, Mohammed stated that they would “tightly and firmly load [the missiles] in our car and bring [them]” for use in the planned attack. Government Trial Ex. 2C. Eleven days later, he told Jaweed he intended to use his portion of the proceeds of the drug sales to purchase a car. Government Trial Ex. 2D Mohammed argues that his statements about buying a car indicate nothing more than that he intended to buy the car for his personal use or to help in his drug trafficking. He claims that his statements cannot be read to support the conclusion of the district court that he was referring to the same car that he said earlier would carry the missiles. Although Mohammed’s objection may show that the record can support alternate interpretations, it is far from proof that the district court’s reading of these conversations is clearly erroneous. See United States v. Erazo, 628 F.3d 608, 611 (D.C. Cir. 2011) (holding that we review factual findings underlying a decision to apply a sentencing enhancement for clear error, and give due deference to the district court’s application of the Guidelines to the facts). Clear error review is exacting: to reverse a district court’s findings of fact “we must be ‘left with the definite and firm conviction that a mistake has been committed.’” Am. Soc’y for the Prevention of Cruelty to Animals [16] v. Feld Entm’t, Inc., 659 F.3d 13, 22 (D.C. Cir. 2011) (quoting Anderson v. City of Bessemer City, 470 U.S. 564, 573 (1985)). Mohammed’s objection does not reach this level of certainty. The district court pointed to specific statements in the record — which Mohammed does not dispute he made — from which it drew plausible inferences. That Mohammed may have intended the car for personal use does not mean he could not also have planned to use the car in the attack, and he identifies no evidence directly contradicting the district court’s conclusion that he did. Especially given the district court’s superior vantage point to make credibility determinations and glean “insights not conveyed by the record,” Gall v. United States, 552 U.S. 38, 51 (2007), we cannot conclude that its findings are clearly wrong.3 V Finally, Mohammed claims that his trial counsel was ineffective because he failed to adequately explore the possibility that evidence was available that would have significantly strengthened Mohammed’s defense. Prior to trial, Mohammed identified for his attorney certain witnesses from his village in Afghanistan who he claimed could bolster his character and impugn Jaweed’s. Mohammed’s attorney admits he did not try to locate or interview any of them. See, e.g., Tr. Status Hr’g 14:1-7, Feb. 25, 2008. Mohammed now claims that these witnesses could have shown that Jaweed had a reputation as a liar and was biased against him. Failing to introduce their testimony prejudiced his defense, Mohammed argues, because [3] In light of this conclusion, we need not consider the district court’s second basis for applying the terrorism enhancement, nor the government’s alternate argument that any error would have been harmless because the district court stated it would have imposed the same sentence without the enhancement. [17] Jaweed was the government’s star witness, and his credibility was central to the prosecution. When advancing an ineffective assistance argument on direct appeal, an appellant must present “factual allegations that, if true, would establish a violation of his [S]ixth [A]mendment right to counsel.” United States v. Poston, 902 F.2d 90 , 99 n.9 (D.C. Cir. 1990). These allegations must satisfy both prongs of Strickland v. Washington, 466 U.S. 668 (1984): deficient representation and prejudice. Id. at 687 . Presented with a colorable claim, we remand for an evidentiary hearing unless the “record alone conclusively shows that the defendant either is or is not entitled to relief.” United States v. Burroughs, 613 F.3d 233, 238 (D.C. Cir. 2010) (quoting United States v. Rashad, 331 F.3d 908, 910 (D.C. Cir. 2003)) (internal quotation marks omitted). We do not “reflexively remand,” United States v. Harris, 491 F.3d 440, 443 (D.C. Cir. 2007), but neither will we hesitate to remand when a trial record is insufficient to assess the full circumstances and rationales informing the strategic decisions of trial counsel, see Massaro v. United States, 538 U.S. 500, 505 (2003). To raise a colorable claim that his trial counsel was deficient, Mohammed must allege errors “so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment.” Strickland, 466 U.S. at 687 . Mohammed’s attorney owed him a “duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary,” id. at 691 , so an allegation that counsel failed to contact potentially vital defense witnesses at Mohammed’s request, without good cause, could rise to this level. The key to Mohammed’s claim is whether his attorney’s less-than-thorough pre-trial investigation was supported by “reasonable professional judgments.” Id.; see also Wiggins v. 18 Smith, 539 U.S. 510, 523 (2003) (noting that courts use an objective standard to assess performance under prevailing professional standards). Pointing to three hearings that took place on May 2, 2008, the government maintains there is enough in the district court record for us to determine whether it was reasonable for Mohammed’s counsel not to search for these potential witnesses in Afghanistan. We disagree. At the ex parte hearing called to consider Mohammed’s request for new counsel, the district court asked Mohammed directly what his witnesses would say at trial. He answered that they would testify that he was not part of the Taliban and that Jaweed was a thief. Tr. Ex Parte Sealed Discussions 23:12-24:3. According to the government, testimony on these points would have been irrelevant to Mohammed’s defense in light of both the government’s stipulation, made at the final hearing of the day, that it would not introduce evidence of Mohammed’s Taliban involvement, as well as Mohammed’s agreement, made in the face of the motion in limine, that he would not to try to impeach Jaweed on this account. Cf. United States v. Moore, 104 F.3d 377, 391 (D.C. Cir. 1997) (“Even had [counsel] located these witnesses, the testimony they allegedly would have provided was tangential at best.”). But the government misses the thrust of Mohammed’s argument. Although Mohammed emphasizes his attorney’s failure to contact potential witnesses, as he did before the district court, on appeal he argues that counsel should have made some effort to learn if the witnesses could undermine Jaweed’s credibility in general, by testifying, for example, that he had a reputation for dishonesty or that he harbored a grudge against Mohammed. The colloquies at the May 2 hearings did not explore that possibility or consider other facts that might be [19] developed by the district court on remand to give fuller context to the decisions counsel made, such as what Mohammed told his attorney while preparing his defense or what his attorney may have uncovered himself during trial preparation. The government argues that Mohammed’s failure to press before the district court this point about Jaweed’s credibility in general shows that these witnesses would not have been able to support his allegation. We think the government infers too much from a colloquy between Mohammed and the court at a hastily convened ex parte hearing on the eve of trial. The question of ineffective assistance looks to the propriety of his attorney’s decision not to follow up on Mohammed’s suggestions to track down possibly helpful witnesses in Afghanistan. If anything, the fact that his attorney noted for the court the possibility that the witnesses could have undermined Jaweed’s testimony more generally shows that he was aware of their potential value to Mohammed’s defense. The government urges us to disregard this statement as mere speculation, but whether counsel did or should have had reason to investigate this hunch before trial is precisely the type of question an evidentiary hearing is designed to explore. The record before us may well be sufficient to judge the reasonableness of not pursuing witnesses who could refute evidence that Mohammed was part of the Taliban or that Jaweed was a thief, but it cannot tell us whether Mohammed’s attorney was justified in not finding out if they had anything else to offer his client’s defense. We do not know “all the circumstances” animating counsel’s strategic decisions from which we could determine whether his failure to pursue this potential line of defense was a reasonable, calculated choice or a mark of deficient performance. See Strickland, 466 U.S. at 691 . Even so, remand would not be needed if the record showed no prejudice to Mohammed from this alleged error. At this [20] stage, Mohammed need not prove actual prejudice, but merely show that the record does not “conclusively establish[] that he could not do so if given the chance.” Rashad, 331 F.3d at 912 . The government argues there is no reasonable likelihood that testimony undercutting Jaweed’s credibility would have altered the verdict because the vast majority of evidence against Mohammed consisted of his own words and actions caught on tape, not Jaweed’s testimony. Mohammed responds that the recordings do not speak for themselves. The government put them into evidence through Jaweed, who explained the recordings and provided context for them over two days of testimony. Mohammed has the better of this argument. The prosecutor frequently stopped the recordings at trial and asked Jaweed to explain them to the jury. For example, Jaweed testified that when Mohammed discussed blowing up mines around “Dagosar” and “the Red Castle” he was referring to government cars, Trial Tr. 46:9-20, May 9, 2008, and that plans to “fire missiles toward the airport,” referenced the Jalalabad airfield, id. at 48:25-49:3. Jaweed’s testimony arguably shaped how the jury understood Mohammed’s words. Without the additional information Jaweed provided that Mohammed was discussing government targets, a juror conceivably could conclude that Mohammed was violent, but not a terrorist as required to convict under § 960a. Errors that have “a pervasive effect on the inferences to be drawn from the evidence” have a greater probability of influencing the verdict, Strickland, 466 U.S. at 695-96 , and Mohammed has raised a colorable claim that his attorney’s failure to introduce evidence challenging Jaweed’s credibility was such an error. The district court is best positioned to answer in the first instance whether this colorable claim rises to the level of actual prejudice. See Massaro, 538 U.S. at 506 (explaining that the district court has an “advantageous [21] perspective” to assess prejudice within the full context of a trial, especially when the same judge from trial presides). Because Mohammed has raised colorable claims under both Strickland prongs and the trial record does not conclusively show whether he is entitled to relief, we remand his claims to the district court to test his allegations further.4 VI For the foregoing reasons, we affirm Mohammed’s judgment of conviction and sentence in all respects, but remand for an evidentiary hearing on his ineffective assistance claim. So ordered. [4] We agree with our concurring colleague that precedent and sound policy mark the district court as the best forum to litigate a claim of ineffective assistance, and we express no view on the merits of Mohammed’s claim. Our discussion seeks only to explain the reasons for our conclusion that the trial record is insufficient to resolve his claim on direct appeal. In most cases, the need for an evidentiary hearing is readily apparent because the trial record contains little of the information necessary to assess trial counsel’s performance. In such cases we may well “owe no special explanation when we remand.” Concurring Op. at 2. But this is not the typical case, because the performance of trial counsel was an issue before the district court. In such an unusual circumstance, we see nothing amiss with explaining why the trial record is insufficient to weigh the merits of a claim of ineffective assistance on direct appeal. In order to respect our charge to avoid a “reflexive[] remand,” we must “interrogate the trial record according to Strickland’s familiar two prongs.” Harris, 491 F.3d at 443 . KAVANAUGH, Circuit Judge, concurring in part and concurring in the judgment: I concur in the judgment and in all but Part V of the Court’s excellent opinion. I write separately with respect to Part V only to express my respectful view that this Court should not ordinarily delve into the merits of an ineffective-assistance claim before the district court has done so. The Supreme Court has stated that “ineffective-assistance claims ordinarily will be litigated in the first instance in the district court, the forum best suited to developing the facts necessary to determining the adequacy of representation during an entire trial.” Massaro v. United States, 538 U.S. 500, 505 (2003). The district court “may take testimony from witnesses for the defendant and the prosecution and from the counsel alleged to have rendered the deficient performance.” Id. For that reason, ineffective-assistance claims arising out of federal criminal cases are most appropriately brought in § 2255 collateral proceedings. (The Supreme Court has said that procedural default rules do not preclude a defendant from bringing an ineffective-assistance claim for the first time in a § 2255 proceeding. See id. at 504.) To be sure, this Court has also permitted ineffective-assistance claims to be raised on direct appeal. 1 But even so, when an ineffective-assistance argument is asserted on direct appeal, our usual practice is to remand the claim to the district court without substantial analysis by this Court of the merits of the claim. See, e.g., United States v. Laureys, 653 F.3d 27, 34 (D.C. Cir. 2011). [1] Our circuit is alone in permitting this procedure. At some point, we perhaps should conform our practice to that of all of the other circuits and require most ineffective-assistance claims to be raised in § 2255 proceedings, not on direct appeal. Cf. Martinez v. Ryan, 132 S. Ct. 1309, 1318 (2012) (“there are sound reasons for deferring consideration of ineffective-assistance-of-trial-counsel claims until the collateral-review stage”). Regardless of whether we take that logical step, however, we should still give the district court the first opportunity to consider such claims. [2] Whether it be in a § 2255 proceeding or on direct appeal, the key procedural principle remains the same: The district court should take the first crack at the merits of ineffectiveassistance claims. Two principles of sound appellate decisionmaking support that district-court-first practice. First, as the Supreme Court has explained, the district court is the forum “best suited to developing the facts.” Massaro, 538 U.S. at 505 . Otherwise, “appellate counsel and the court must proceed on a trial record not developed precisely for the object of litigating or preserving the claim and thus often incomplete or inadequate for this purpose.” Id. at 504-05 . Second, by remanding to the district court as a matter of course when an ineffective-assistance claim is raised on direct appeal, we avoid wasting scarce resources as appellate counsel (and judges) fruitlessly and pointlessly squabble over ineffectiveassistance claims based on an incomplete record. We have acknowledged that the court of appeals can resolve an ineffective-assistance issue in the first instance when the record “conclusively” shows that the defendant either is or is not entitled to relief. United States v. Rashad, 331 F.3d 908, 911 (D.C. Cir. 2003). But given the fact-bound nature of ineffective-assistance claims, that exception arises only rarely. If there is any doubt or difficulty, if it is not obvious from the face of the record whether relief is warranted, the appropriate course is simply to remand. Applying those principles to this case, I do not see the need for the Court’s detailed analysis of Mohammed’s ineffective-assistance claim. We owe no special explanation when we remand an ineffective-assistance claim. We owe a special explanation only in the rare situations when we resolve the ineffective-assistance claim here at the appellate level. In this case, I would remand the ineffective-assistance [3] claim to the district court without the lengthy evaluation of the claim’s merits.