Five defendants — appellants Baugham and Wells and three others, Honesty, White, and James Nelson, Jr.1 — were tried together for conspiracy to distribute and to possess with intent to distribute 50 grams or more of cocaine base, in violation of 21 U.S.C. § 846. Each of the five defendants was also tried in the same proceeding on one or more substantive offenses, each involving drugs, guns, or both. Honesty, White, and James Nelson, Jr., were acquitted on all counts. Baugham was convicted on the conspiracy count, plus two others: distributing cocaine base, in violation of 21 U.S.C. §§ 841(a)(1) and (b)(1)(C) and 18 U.S.C. § 2, and possessing 5 grams or more of cocaine base with intent to distribute, in violation of 21 U.S.C. §§ 841(a)(1), (b)(1)(B)(iii) and 18 U.S.C. § 2. Wells was also convicted on the conspiracy count, plus two others: distributing 50 grams or more of cocaine base, in violation of 21 U.S.C. § 841(a)(1) and (b)(1)(A)(iii) and 18 U.S.C. § 2, and possessing a firearm in furtherance of a drug trafficking offense (i.e., the drug conspiracy), in violation of 18 U.S.C. § 924(c)(1).

Baugham and Wells attack their convictions on a variety of grounds; the only ones meriting discussion in a published opinion are claims of insufficiency of evidence of conspiracy and of a fatal variance between the conspiracy alleged and the proof at trial. They also challenge their sentences. We affirm the convictions but vacate the sentences and remand for resentencing.

Baugham argues that there was insufficient evidence that he conspired with any of the defendants, cooperators, or informants. It is unclear whether Wells also mounts a sufficiency challenge, but the government reads his brief as doing so and therefore presents what it contends is evidence sufficient to support both appellants' conspiracy convictions. Brief of Appellee at 29-37, esp. 34, 36. Given that this court in any event has the power to notice a plain error sua sponte, Silber v. United States, 370 U.S. 717, 718, 82 S. Ct. 1287, 8 L. Ed. 2d 798 (1962), and assuming in Wells's favor that our usual deference to the jury verdict is no greater when the plain-error rule applies, see United States v. Spinner, 152 F.3d 950, 956 (D.C.Cir. 1998), we think it appropriate, under these circumstances, to subject his conspiracy conviction to the same scrutiny as Baugham's.

Evidence is sufficient if, when viewed in the light most favorable to the government, it would permit a rational jury to find the elements of conspiracy beyond a reasonable doubt. United States v. Graham, 83 F.3d 1466, 1471 (D.C.Cir. 1996). The drug conspiracy statute, 21 U.S.C. § 846, dispenses with the usual requirement of an overt act and requires only an agreement to commit any offense(s) defined in the subchapter, United States v. Shabani, 513 U.S. 10, 16-17, 115 S. Ct. 382, 130 L. Ed. 2d 225 (1994) — in this case, distribution of, or possession with intent to distribute, 50 grams or more of "cocaine base," 21 U.S.C. §§ 841(a)(1), (b)(1)(A)(iii).

The sufficiency and variance issues interact with each other here in a rather complex way. In United States v. Brisbane, 367 F.3d 910 (D.C.Cir.2004), we addressed the "cocaine base" element of the offense, holding that a conviction premised on "cocaine base" under 21 U.S.C. § 841 cannot stand unless the evidence establishes that the cocaine at issue was crack or that it was smokable; we left unresolved whether proof of smokability alone would suffice. Id. at 914. Appellants did not raise the Brisbane problem below, so we review for plain error under Federal Rule of Criminal Procedure 52(b). Thus, the convictions cannot stand if (1) there is error (2) that is plain and (3) that affects substantial rights, and (4) we find that the error "seriously affects the fairness, integrity or public reputation of judicial proceedings." United States v. Olano, 507 U.S. 725, 732, 113 S. Ct. 1770, 123 L. Ed. 2d 508 (1993) (citations, internal quotation marks, and brackets omitted).

Here, if the evidence turns out to be sufficient to support the conviction even when the statute is read in accordance with Brisbane, there is no error at all as to sufficiency. Because of all hands' failure to anticipate Brisbane, the record is relatively weak on whether either the conspiracy alleged in the indictment, or even a Baugham-Wells conspiracy argued by the government as a fallback, actually involved 50 grams of cocaine base as defined in Brisbane. But the evidence, as we shall soon see, is quite abundant for a showing that Baugham and Wells each conspired with cooperating witness Earl Nelson to distribute far more than 50 grams of crack.

Because the parties' briefs hadn't addressed the relation between Brisbane and the sufficiency and variance issues, we ordered a second round of briefing, putting to the parties the questions (among others) whether the evidence was adequate to show Baugham-Earl and Wells-Earl conspiracies on the scale of 50 grams of cocaine base as defined in Brisbane and whether reliance on those conspiracies would mean that the variance caused appellants harm justifying reversal. Having studied the second round of briefs, we proceed to those two issues.

Cooperating witness Earl Nelson testified that he purchased an ounce (i.e., about 28 grams) of "crack cocaine" from Baugham "more than twenty or thirty" times and that he purchased even more ounces of crack cocaine from Wells.

A sale of drugs does not, however, per se establish a conspiracy between seller and buyer to distribute such drugs, or to possess them with intent to distribute. In drawing the distinction between a conspiracy and a non-conspiratorial buyer-seller relationship, the Supreme Court and this court have considered a variety of factors: the seller's knowledge of the buyer's illegal purpose (e.g., to re-sell the drugs) and of any larger organization designed to further that purpose; the seller's intent to further the buyer's illegal purpose; the duration and regularity of the dealings; the quantity of drugs sold; the importance of the particular buyer to the particular seller and vice versa; and either party's special efforts to get the other party to transact with him (including extension of credit) or to make the entire operation run successfully. See Direct Sales Co. v. United States, 319 U.S. 703, 711-12 & n. 8, 63 S. Ct. 1265, 87 L. Ed. 1674 (1943); United States v. Thomas, 114 F.3d 228, 245 & n. 5 (D.C.Cir.1997); United States v. Baylor, 97 F.3d 542, 543-44, 547 (D.C.Cir.1996); United States v. Sobamowo, 892 F.2d 90, 94-95 (D.C.Cir.1989); United States v. Morris, 836 F.2d 1371, 1373-74 (D.C.Cir.1988). While the seller's knowledge of the buyer's illegal purpose is necessary to conviction, Direct Sales, 319 U.S. at 712 & n. 8, 63 S. Ct. 1265, and multiple huge sales (much larger than those here) may be sufficient, United States v. Childress, 58 F.3d 693, 714 (D.C.Cir.1995), the cases otherwise say little about how the various factors are to be weighed.

We conclude that Earl's purchases from each of the appellants separately fulfill enough of the indicia of conspiracy to support their convictions. We do not decide whether the evidence supports the single broad horizontal-and-vertical conspiracy alleged in the indictment, nor any part of it (e.g., the Baugham-Wells conspiracy identified by the government), aside from the Earl-Baugham and Earl-Wells agreements. We save the variance issue for the next section.

Earl plainly intended to distribute the crack he bought from each appellant. When asked about his general practice when purchasing ounces from Baugham, Earl testified that the purchase price was $1000 and that if the entire quantity were sold at retail (i.e., "[i]f you bag up an ounce"), the gross revenue would be $2000. In fact, however, Earl said his profit if he "did good" was only $200 or $300, since he exchanged some of the crack for sex, sometimes didn't get the full price from his retail customers, and smoked some of the crack himself (the only non-distributed portion). Thus, Earl must have typically distributed around half or more of each ounce he purchased from Baugham.

The jury could rationally infer that this testimony also applied to the ounces that Earl purchased from Wells. Earl gave no reason to think his practice differed depending on his supplier. Moreover, immediately after giving the testimony on Baugham cited above, Earl shifted seamlessly into comments that applied to all of his suppliers, suggesting he didn't mean his testimony to be specific to Baugham.

Further, there is evidence that each appellant knew of Earl's intent to distribute and sought to further the distribution. Testifying as to his general practice when purchasing ounces from Baugham, Earl said that he could sell an ounce in a day and a half if he worked hard. He said he would sometimes take breaks in the crack house until whoever was supplying him — and here he named Baugham and Wells as examples — would "come in the crack house hollering what you doing here in it or why you in here and run me out because you can't make no money in here."

In addition, Earl said that Baugham and Wells were aware of his distribution activities during earlier phases of his career during which he was buying sub-ounce quantities, and that they then kept tabs on his selling. On this basis, the jury had every reason to infer that appellants remained aware of Earl's distribution activities once they began selling him even larger quantities.

There is also evidence that Earl dealt with each appellant regularly and over a long period. Earl testified that his drug-dealing relationship with Baugham began in 1993, and audiotape evidence confirms that he bought an ounce from him in 2000. Earl said that, over the course of his relationship with Baugham, he initially received sets of tiny "dime bags" at least two to three times per week; next "eight-balls" (each one-eighth of an ounce) one or two times per week, "if [he] did that many"; next quarter-ounces, though Earl didn't say how frequently; and finally ounces, at least two or three times per week, for a total of over twenty or thirty. As for Wells, Earl said he began dealing drugs for him in 1993 and continued (with a one-year interruption) over a ten-year period, which would cover most of the alleged period of the conspiracy (1992-2000). Earl said that when he was buying ounces from Wells the frequency was at least two or three per week, for a total of more than twenty or thirty.

As we mentioned, credit arrangements can be a significant factor supporting a buyer-seller conspiracy; this is true even if they apply to only some of the transactions. See United States v. Melendez, 401 F.3d 851, 854 (7th Cir.2005); United States v. Burroughs, 830 F.2d 1574, 1581 (11th Cir.1987). Earl testified that when he bought ounces from Baugham, he received them on credit sometimes, depending on whether he had the cash and whether Baugham felt generous. When asked about his general practice after obtaining an ounce, Earl responded that he would only "pay him [i.e., Baugham] his thousand" after he'd smoked some of the crack and distributed the rest, suggesting that sales on credit were a familiar part of their relationship.

As to Wells, Earl said some of their dealings were on credit in the early part of his career, when he sold very small quantities. He also testified that people who sold to him on credit would investigate stories he told of his stash being stolen; when asked who these investigating people were, Earl replied, "[p]eople that fronted me" and named Wells (and Baugham) as examples. Further, audiotape evidence shows that, when Baugham and Wells jointly sold an ounce to Earl in May 2000, they agreed that Earl would pay $1000 up front, that Baugham would deliver the drugs later in the day, and that Earl would pay an additional $200 at some unspecified later time. Earl testified that he trusted Wells to follow through on the delivery because "we got a history, so I could trust him. I had trust in him just like he had it in me . . . . But I'm saying if I gave him my money I know it was good. I never had a problem with him before, so I ain't — I knew I wasn't going to have [one] with [him] then. He was going to deliver." As to the deferred payment of $200, Earl added that "I told [Wells] I don't have nothing but the thousand. You know I'm good, I owe you shit and I ain't try to owe you. If I give you thousand, let me give you the two tomorrow. So he knew it was good, so he went for it. It was on him just like it was on me. You had to put trust or something in each other to make the deal go through." To be sure, this testimony concerns both deferred payment and deferred delivery, whereas our cases tend to focus on the former. Yet both indicate "a level of mutual trust," which at least one other circuit has reasonably identified as an indicator of a buyer-seller conspiracy. See Melendez, 401 F.3d at 854.

There is also the issue of the buyer's importance to the seller and vice versa. Earl had no monopoly on distribution services; many people sold drugs at retail in his selling area. The government certainly did not allege that Baugham or Wells used Earl exclusively, and there's no reason to think either appellant was especially dependent on Earl, though the mutual build-up of trust between them presumably would have meant there was some cost to replacing him.

As to the parties' awareness of a larger organization, video and audiotape evidence and the testimony of Earl, a police detective, and a DEA chemist establish Baugham and Wells's cooperation to sell a substantial quantity of crack to Earl in May 2000.

Overall, enough of the indicators of a buyer-seller conspiracy are present to sustain the convictions. The trial court charged the jury on the issue, in terms that correctly stated the law, and to which appellants have raised no objection in either the first or second round of briefing.

Thus, the government presented, at the very least, evidence sufficient to convict each appellant of a conspiracy with Earl to distribute, or to possess with intent to distribute, 50 grams of cocaine base under 21 U.S.C. § 846.

What we've just said, of course, doesn't resolve appellants' claims that the evidence at trial was insufficient to prove the single conspiracy alleged in the indictment (involving all five defendants plus various co-conspirators) and that the variance between indictment and proof warrants reversal. In response to those claims, the government argues energetically that the proof was sufficient to show the five-defendant-plus conspiracy. We think the proof of such a conspiracy exiguous at best, and as to the narrower Baugham-Wells conspiracy, exiguous on the assumption that Brisbane applies. But we need not go into the matter, for even if the proof suffered from this deficiency — even if it showed three narrow and separate conspiracies, one between Earl and Baugham, another between Earl and Wells and yet another (as appellants contend) among Honesty, James Nelson, Jr., and others — such a variance does not warrant reversal if it was harmless. We conclude that it was.

We first address the standard of review and the burden of persuasion, and then apply them to the record.

We start our harmless error analysis by addressing the tension between general statements about harmless error review of constitutional errors and actual judicial treatment of the sort of variance alleged here.

In their opening briefs, appellants assert only one type of harm arising from variance: "transference of guilt," especially jury confusion and improper reliance on testimony against one defendant to convict another (possibly inadmissible against the latter defendant under the hearsay rule). Further, in response to this court's questions about evidence relating to Earl Nelson, Wells claims that affirmance on the basis of his agreement with Earl would be unfair because he lacked notice of this theory at trial.

The proper standard of review for the type of variance claimed here is the conventional one, articulated in Kotteakos v. United States, 328 U.S. 750, 776, 66 S. Ct. 1239, 90 L. Ed. 1557 (1946), i.e., whether the error had a "substantial and injurious effect or influence in determining the jury's verdict," rather than the one set forth in Chapman v. California, 386 U.S. 18, 24, 87 S. Ct. 824, 17 L. Ed. 2d 705 (1967), requiring that for constitutional errors the government must normally show lack of prejudice beyond a reasonable doubt. The Court decided Chapman, significantly, against a well-nigh universal background understanding that all constitutional errors, at least in federal court, required reversal regardless of any specific reason to infer an impact. 5 LAFAVE, ISRAEL & KING, CRIMINAL PROCEDURE § 27.6(c) (2d ed. 1999 & Supp.2006).

Some divergences between indictment and proof (we use the neutral term "divergence" advisedly, to encompass both "constructive amendments" and "variances," see 4 LAFAVE ET AL., supra § 19.6(c)) plainly have a constitutional dimension. They surely relate to an accused's Fifth Amendment right not to be tried for a felony "unless on a presentment or indictment of a Grand Jury" and to his Sixth Amendment right "to be informed of the nature and cause of the accusation." Less obviously, and perhaps less realistically in light of Blockburger v. United States, 284 U.S. 299, 52 S. Ct. 180, 76 L. Ed. 306 (1932), see 4 LAFAVE ET AL., supra, § 19.6(b) at 808, a divergence might expose the defendant to a risk of double jeopardy in violation of the Fifth Amendment. See Berger v. United States, 295 U.S. 78, 82, 55 S. Ct. 629, 79 L. Ed. 1314 (1935) (noting possibilities of surprise and double jeopardy).

Despite these attributes of divergence errors and the general principle said to govern constitutional error, no decision of the Supreme Court (nor indeed of any other federal appellate court that we can discover) has applied the "reasonable doubt" variant of harmless error, much less automatic reversal, to the kind of divergence claimed here — a charge of one conspiracy but proof of more than one. Berger itself, identifying the possible harms as deprivation of notice (which sounds like the Sixth Amendment right to be informed of the charge) and double jeopardy, straightforwardly applied the prejudice requirement stated in the then-prevailing version of today's harmless error statute (28 U.S.C. § 2111; see also FED. R. CRIM. P. 52(a)). The Court gave no hint that applying the standard might breach a pre-existing norm of automatic reversal. Given that Berger found conventional harmless error analysis appropriate for the divergence at issue despite a background assumption of automatic reversal for constitutional errors, it follows that such analysis remains appropriate in the post-Chapman world, where the background assumption is less demanding for the government than prior to Chapman.

Kotteakos is of course consistent with this conclusion. There the government tried 19 defendants (of whom six were dropped before the case went to the jury) as alleged members of one conspiracy, 328 U.S. at 753 & n. 4, 66 S. Ct. 1239, while the proof instead showed at least eight separate ones, id. at 754-55, 66 S. Ct. 1239 The harm claimed was the risk of "transference of guilt from one [defendant] to another across the line separating conspiracies, subconsciously or otherwise." Id. at 774, 66 S. Ct. 1239; see also id. at 766-67, 769-75, 66 S. Ct. 1239. Finding the risk substantial, the Court reversed, id. at 774, 776, 66 S. Ct. 1239, but as in Berger it applied the harmless error statute and made no reference to reasonable doubt or automatic reversal, id. at 757-76. This flowed naturally from appellants' exclusive reliance on transference of guilt as the harm, since it is not linked to any specific constitutional requirement. When it alone is in question, the error is perhaps best understood as derived from common law, as asserted in one leading treatise, 5 LAFAVE ET AL., supra § 27.6(b) at 940 n. 35, or from the rules governing joinder, which Kotteakos itself offered as an alternative basis for the holding, 328 U.S. at 774-75, 66 S. Ct. 1239, and which one circuit has argued (in dictum) are the true basis of the "multiple conspiracy doctrine," United States v. Sutherland, 656 F.2d 1181, 1190 n. 6 (5th Cir.1981) (citing FED. R. CRIM. P. 8(b)).

Although the divergence between one and many conspiracies has constitutional overtones, Berger itself spoke of the potential harm in very mild terms. First, it noted that it was settled that, where the indictment claims a conspiracy of "several persons" and the proof was sufficient only for "some," "the variance is not material." 295 U.S. at 81, 55 S. Ct. 629. Then:

If the proof had been confined to that conspiracy [one conspiracy, but inadequate against one or more of the accused], the variance, as we have seen, would not have been fatal. Does it become so because, in addition to proof of the conspiracy with which petitioner was connected, proof of a conspiracy with which he was not connected was also furnished and made the basis of a verdict against others?

Id. Answering its own (perhaps rhetorical) question, the Court proceeded to apply a conventional harmless error analysis. Id. at 81-84, 55 S. Ct. 629. Cf. Brecht v. Abrahamson, 507 U.S. 619, 631-32 & n. 7, 113 S. Ct. 1710, 123 L. Ed. 2d 353 (1993) (alluding to Kotteakos as a paradigmatic application of the then-prevailing harmless error statute, which is substantively identical to the current one, 28 U.S.C. § 2111, so far as the harmless error standard itself goes).

Berger and Kotteakos contrast sharply with the more acute divergences (labeled "constructive amendments") that lead to reversal automatically, without a specific inquiry as to harm. Thus in Stirone v. United States, 361 U.S. 212, 80 S. Ct. 270, 4 L. Ed. 2d 252 (1960), the defendant was charged with unlawfully obstructing shipments of sand into Pennsylvania, while at trial the prosecutor offered evidence not only of that but also of the defendant's obstruction of shipments of steel out of Pennsylvania. Id. at 213-14, 80 S. Ct. 270. The jury could have convicted on the basis of shipments of a different commodity and in a different direction from the ones stated the indictment. Id. at 218-19, 80 S. Ct. 270. Compare United States v. Miller, 471 U.S. 130, 105 S. Ct. 1811, 85 L. Ed. 2d 99 (1985); see also 1 WRIGHT, FEDERAL PRACTICE AND PROCEDURE: CRIMINAL § 127 (3d ed. 1999 & Supp.2004); 4 LAFAVE ET AL., supra § 19.6(c).

The harms claimed here are like the ones at issue in Berger (surprise) and Kotteakos (transference). We therefore apply the conventional harmless error standard, i.e., error is harmful if it had a "substantial and injurious effect or influence in determining the jury's verdict." United States v. Powell, 334 F.3d 42, 45 (D.C.Cir.2003) (quoting Kotteakos, 328 U.S. at 776, 66 S. Ct. 1239). The question is "whether the guilty verdict actually rendered in this trial was surely unattributable to the error," Sullivan v. Louisiana, 508 U.S. 275, 279, 113 S. Ct. 2078, 124 L. Ed. 2d 182 (1993); see also United States v. Maddox, 156 F.3d 1280, 1283-84 (D.C.Cir.1998); United States v. Cunningham, 145 F.3d 1385, 1394 (D.C.Cir.1998).2 This inquiry leads us to ask how correction of the particular error at issue might have altered either a defendant's strategy or the jury's thinking and whether such alteration (if any) might have led to a different verdict in light of all the evidence. An error whose correction could neither have materially enhanced a defendant's ability to present a valid defense, nor have had any plausible effect on jury reasoning, cannot be prejudicial, even if the totality of the government's evidence on a given charge is less than overwhelming.

It is firmly established that the government bears the burden of showing the absence of prejudice in harmless error cases generally, whether decided under Kotteakos, see Olano, 507 U.S. at 734, 741, 113 S. Ct. 1770; Powell, 334 F.3d at 45; United States v. Smart, 98 F.3d 1379, 1390 (D.C.Cir.1996), or Chapman, see Satterwhite v. Texas, 486 U.S. 249, 256, 108 S. Ct. 1792, 100 L. Ed. 2d 284 (1988).

Against this background, variance law presents an apparent anomaly. The conspiracy variance cases in our circuit have generally said that the defendant bears the burden of showing prejudice. United States v. Mathis, 216 F.3d 18, 23 (D.C.Cir. 2000); United States v. Gaviria, 116 F.3d 1498, 1516 (D.C.Cir.1997); United States v. Graham, 83 F.3d 1466, 1471 (D.C.Cir. 1996); United States v. Childress, 58 F.3d 693, 709 (D.C.Cir.1995); United States v. Tarantino, 846 F.2d 1384, 1391 (D.C.Cir. 1988). See also United States v. Anderson, 39 F.3d 331, 348 (D.C.Cir.1994) (assuming the evidence supported multiple conspiracies rather than the single one alleged, no reversal was in order because the prejudice requirement was not fulfilled), rev'd in part on other grounds, 59 F.3d 1323 (D.C.Cir.1995) (en banc). Similar statements about how variance claims should generally be treated can be found in the case law of every circuit with criminal jurisdiction. E.g., United States v. Stigler, 413 F.3d 588, 592 (7th Cir.2005); United States v. Salmonese, 352 F.3d 608, 621-22 (2d Cir.2003); United States v. Herrera, 289 F.3d 311, 318-19 (5th Cir.), vacated pending review en banc 300 F.3d 530 (5th Cir.), reinstated in relevant part, 313 F.3d 882, 885 (5th Cir.2002) (en banc); United States v. Lopez-Arce, 267 F.3d 775, 781 (8th Cir.2001); United States v. Crozier, 259 F.3d 503, 519 (6th Cir.2001); United States v. Moore, 198 F.3d 793, 795-96 (10th Cir.1999); United States v. Candelaria-Silva, 166 F.3d 19, 40 (1st Cir.1999); United States v. Romer, 148 F.3d 359, 369 (4th Cir.1998), overruled on other grounds as recognized in United States v. Strassini, 59 Fed.Appx. 550, 552 (4th Cir.2003); United States v. Balter, 91 F.3d 427, 441 (3d Cir.1996); United States v. Starrett, 55 F.3d 1525, 1553 (11th Cir.1995); United States v. Homick, 964 F.2d 899, 907 (9th Cir.1992). Indeed, we have found no circuit decision to the contrary.

A possible explanation might be that the prejudice from a variance is an element of the violation itself, in the same way that the Supreme Court conceives of prejudice as an element of the violation of the right to effective assistance of counsel, Strickland v. Washington, 466 U.S. 668, 692, 104 S. Ct. 2052, 80 L. Ed. 2d 674 (1984), and of the right to prosecutorial disclosure of certain evidence, Kyles v. Whitley, 514 U.S. 419, 435-36, 115 S. Ct. 1555, 131 L. Ed. 2d 490 (1995). See also 5 LAFAVE ET AL., supra § 27.6(d) at 947-48 & n. 74. On that view, it makes sense to require the defendant to show prejudice. E.g., Strickler v. Greene, 527 U.S. 263, 291, 119 S. Ct. 1936, 144 L. Ed. 2d 286 (1999) (in prosecutorial nondisclosure case, "petitioner's burden is to establish a reasonable probability of a different result"); Strickland, 466 U.S. at 687, 104 S. Ct. 2052 ("defendant must show that [counsel's] deficient performance prejudiced the defense"). (Though Strickland, Kyles, and Strickler are all habeas cases, it seems probable that the formula is the same for all ineffectiveness and nondisclosure claims, see, e.g., 5 LAFAVE ET AL., supra § 27.6(d) at 947-48 & n. 74.) But this theory seems to run aground on the fact that both Berger, 295 U.S. at 81-84, 55 S. Ct. 629, and Kotteakos, 328 U.S. at 757-76, 66 S. Ct. 1239, rely heavily on the harmless error statute, which would be unnecessary if, in the absence of prejudice, there were no violation at all. Cf. Kyles, 514 U.S. at 435-37, 115 S. Ct. 1555 (discussing prejudice element of prosecutorial nondisclosure claim without reference to harmless error statute or rule); Strickland, 466 U.S. at 691-96, 104 S. Ct. 2052 (same for ineffective assistance of counsel).

We need not resolve this conflict here. On this record, the government prevails even if it bears the burden. In making that determination, we are guided by Sullivan's directive to focus on whether the error might have accounted for the outcome.

We first consider the issue of transference of "guilt" from one defendant to another. The most obvious ways in which the inclusion of four unnecessary co-defendants (i.e., the three acquitted defendants plus the other appellant) might have affected the jury's reasoning process are through (1) evidence against either appellant that wouldn't have been admissible in the absence of the four co-defendants, and (2) jury confusion.

Here, for each of the convictions, the government's case rested on evidence that would have been admitted even if Baugham and Wells had each been tried alone. To show this, we consider each conviction in turn.

First, as to conspiracy, we have already analyzed the evidence that Baugham and Wells each conspired with Earl, whose testimony was admitted directly against each appellant.

Baugham's conviction on the count of distributing cocaine base rested on a videotape of Baugham handing a clear plastic envelope to one Darryl Young, and the testimony of the videotaping detectives, who after witnessing the hand-off contacted other officers and asked them to follow Young. A chemist's report identifies as cocaine base the contents of a plastic envelope found on an entertainment center in the apartment into which Young fled. Although some of this evidence is hearsay, arguably non-admissible hearsay, its admission didn't depend on the inclusion in the indictment of the other four defendants, and Baugham has made no challenge to the sufficiency of the evidence on this count.3

Baugham's conviction on the count of possessing 5 grams or more of cocaine base with intent to distribute was equally independent of evidence against the other defendants. It rested on Metropolitan Police Department ("MPD") detective Witkowski's testimony that: he had seen Baugham driving a certain car "on many occasions"; that he acquired a warrant to seize the car; that Baugham, upon encountering Witkowski and other officers, fled in that same car; that Witkowski and other officers located Baugham soon afterward, getting out of the car; that they seized the car; and that Witkowski later participated in a search of the car that produced, from under the back seat, numerous small ziploc bags each containing a white rock substance. A DEA chemist later testified that the substance was crack cocaine. The evidence cited by the government does not support the 5-gram amount; but Baugham has in no way challenged the sufficiency of the evidence underlying this conviction, and we can conceive of no way in which trying Baugham together with his four co-defendants could have made the jury more likely to convict on this count.

Wells's conviction on the count of distributing 50 grams or more of cocaine base rested on testimony of cooperator Perry Nelson that he paid Wells $1900 in exchange for 62 grams of crack; a videotape of Perry entering Wells's apartment building; an audiotape (from a wire worn by Perry) of the transaction itself; and testimony of Witkowski that he received the contraband from Perry immediately after the transaction, that it was a "white rock substance," that he had it tested by the DEA, and that it came to 59 grams.

Finally, Wells's conviction on the count of possessing a firearm in furtherance of a drug trafficking conspiracy rested on the MPD's seizure from Wells's apartment — within four hours of his sale to Perry Nelson described above — of four pistols (three semi-automatic and at least three loaded), ammunition, $6000 cash, a gram scale, a lease agreement between Wells and the apartment owner, and Wells's driving license. The seizure was described at length in the testimony of Witkowski, who supervised it. It was also captured on videotape. Although no drugs were recovered during the seizure, the audiotape of Wells's prior sale to Perry recorded Wells's statement that he was selling Perry all the drugs he had left.

Thus the government has pointed to evidence underlying the three counts against each defendant (the conspiracy count in common, the two separate counts for each) that is in no way dependent on the fact that the five men were tried together.

The only specific evidence pointed out by either appellant that wouldn't have been admitted had each been tried alone is the testimony of one Paula Spriggs. Baugham, in an effort to show that his operation was unrelated to that of Wells, called Spriggs, who said she had purchased drugs from Baugham and then testified as follows:

Q. I had asked you whether or not you had ever purchased drugs from Michael Wells?

A. Yes.

Q. And as between Michael Wells and Reginald Baugham's drugs, whose drugs, if either, had more flavor?

A. Reginald Baugham.

Wells argues that this testimony interfered with his defense of general denial and was therefore prejudicial. We think it was cumulative and therefore not prejudicial. It was brief and extremely general, in contrast to the extensive and far more specific testimony of Earl and Perry Nelson as to Wells's longtime drug-dealing; plus the videotape and audiotape evidence (coupled with the testimony of Earl, a police detective, and a DEA chemist) that Wells made a sale of a substantial quantity of crack to Earl; plus the videotape and audiotape evidence (coupled with the testimony of Perry and a police detective) that Wells sold 59 grams to Perry. Wells contends that the Spriggs testimony was uniquely damaging in that it was elicited by Baugham, his co-defendant and brother. But the identity of the party eliciting Spriggs's testimony does not outweigh the fact that the marginal significance of the information she provided was small. Drawing an analogy to a related doctrine, we note that inconsistent defenses do not per se justify severance, even when co-defendants "point the accusing finger at one another." United States v. Moore, 104 F.3d 377, 384 (D.C.Cir.1997).

Conceivably the 3600-page record contains some additional testimony or exhibits relating to these counts that won admission only because of the broader charge, but appellants haven't pointed us to one iota of such evidence. We are confident, then, that the evidence leading to appellants' convictions would have reached the jury even if each had been tried alone.

Apart from admissibility of otherwise inadmissible evidence, improper transference of guilt might arise from disparities in the weight or type of evidence against the appellants and others, United States v. Badru, 97 F.3d 1471, 1475 (D.C.Cir.1996) (addressing a similar problem in the context of misjoinder), or from the presence of "shocking or inflammatory evidence" that "came in against co-defendants," United States v. Alessi, 638 F.2d 466, 475 (2d Cir.1980). Here the acquittal of the other three defendants indicates no such risk with respect to them, and appellants point us to nothing of the kind. As between Baugham and Wells, the government's cases against the two men were not especially disparate; each was aimed at showing a longtime drug-dealing career, involving the use of guns, via testimony of law enforcement officers and cooperators and video and audiotape evidence of sales.

Kotteakos, 328 U.S. at 766-67, 772-73, 66 S. Ct. 1239, also pointed to the risk of transference that might arise from the sheer number of co-defendants and resulting jury confusion. See also Mathis, 216 F.3d at 25; Gaviria, 116 F.3d at 1533. But the number tried here is hardly of that dimension. Our variance cases have twice approvingly cited Alessi, 638 F.2d at 475, noting both times that the court there found that ten defendants were a "sufficiently small" number to enable the jury to give each individual consideration. Gaviria, 116 F.3d at 1533; Anderson, 39 F.3d at 348.

Finally, we've noted before that "the danger of spillover prejudice is minimal when the Government presents tape recordings of individual defendants," since it's easier, other things being equal, for a jury to match such evidence to individual defendants. Gaviria, 116 F.3d at 1533; see also Mathis, 216 F.3d at 25; Anderson, 39 F.3d at 348. Here, audio and video recordings helped show Baugham and Wells's joint sale to Earl Nelson and Wells's sale to Perry Nelson, and a video recording helped show Baugham's distribution to Young.

As in claims of misjoinder, a variance might sometimes injure defendants because of conflicts between their defenses. Apart from Wells's complaint about Spriggs's testimony, discussed above, appellants make only one claim along these lines. Baugham notes that cooperating witnesses Vincent McSwain, Earl Nelson, Perry Nelson, and James Nelson, Sr., are relatives of co-defendants Honesty and James Nelson, Jr. He suggests (very briefly) that the cooperators testified so as to "minimize the misconduct of their relatives . . . while lying about" how they sold crack for Baugham. Wells adopted this same reasoning at oral argument, again with little elaboration. Oral Arg. Tr. at 6. There are two problems with the theory. First, of the four relative-cooperators, one (McSwain) implicated his relative Honesty more than he did appellants, and another (Perry Nelson), while giving testimony that was overall more damaging to appellants than to his relatives, said that he received very significant amounts of crack from McSwain, who, in turn, admitted he was supplied by Honesty. Thus, the record offers relatively little support for appellants' theory. Second, and perhaps more telling, even if the cooperators' sympathy for their relatives inclined them against appellants, it's unclear how separate trials for the two groups would have helped. Presumably the cooperators agreed to testify in the hopes of more favorable sentencing. These incentives would have remained the same regardless of whether the trials were separate or combined. One can, of course, hypothesize a chain of reasoning supportive of appellants' claim. Suppose that (1) despite their pleas, the cooperators would have liked to go easy on everybody, but (2) the combined trial focused them especially on securing a conviction in that trial, and (3) they therefore directed the jury's wrath away from their relatives and toward appellants. The scenario doesn't seem to us sufficiently plausible to refute the government's showing.

As to "transference of guilt," then, the government has offered convincing reason to believe that any error had no "substantial and injurious effect or influence in determining the jury's verdict." Kotteakos, 328 U.S. at 776, 66 S. Ct. 1239. In reaching this conclusion, we do not rely on its suggestion that the court's charge required the jury to acquit of conspiracy unless it found the "single conspiracy charged in the indictment." Cf. Mathis, 216 F.3d at 25. The trial court had given such a charge in Kotteakos; Justice Douglas, citing Judge Hand's opinion for the Second Circuit, noted that it was error, but an error favoring defendants. See 328 U.S. at 778-79, 66 S. Ct. 1239 (Douglas, J., dissenting). Though favoring the defendants, its role is quite obscure. As the jury acquitted the other three charged conspirators, we must infer that it disregarded that portion of the charge (assuming the passage meant what the government claims). The passage cited gives the government no extra mileage.

Having disposed of the transference issue, we now turn to Wells's claim in his supplemental briefs that his defense suffered from lack of notice that he might be convicted on the basis of his dealings with Earl.

In fact Wells received reasonable notice. The indictment, though not referring to Earl by name, did refer to "other persons known and unknown to the grand jury" and alleged that Baugham and Wells, specifically, "distributed quantities of cocaine base . . . to other drug dealers" in the very neighborhood where Earl said he dealt with appellants. While the government's opening statement referenced a large number of alleged conspirators, it made clear that their "common plan" was "to sell crack cocaine to keep the customers there [i.e., in the neighborhood], to keep the supply going so that everyone can make money"; that the "suppliers" were Baugham, Wells, and Honesty; that the "redistributors . . . who these wholesale suppliers sold to" included Earl Nelson; and that all the conspirators were "acting as agents for each other." Further, the opening statement made four distinct explicit references to Earl's distribution activities on behalf of both Wells and Baugham. At trial, Earl gave more than a full day of testimony, and was cross-examined by counsel for Wells and Baugham. So far as appears, neither Wells nor Baugham ever suggested here or in the district court that he was ambushed by this testimony, and it's hard to imagine how either could have been. In its summation, the government pointed to evidence supporting the claims of vertical conspiracy made in its opening statement, emphasizing that the "resellers" were a "really necessary" part of the selling effort, naming Earl twice as a reseller, and describing some of Earl's distribution activities on behalf of Baugham and Wells.

Nor does Wells explain what difference greater foreknowledge of the ultimate basis for affirmance would have made. He says he would have argued differently as to "the prejudicial evidence presented by Baugham," presumably meaning the Spriggs testimony. But he gives no specifics, and given that the testimony itself was not prejudicial, we cannot imagine how a different treatment of it at trial would have made a difference to the outcome.

Wells also says he would have requested different jury instructions. But the district court gave an instruction on the buyer-seller exception to conspiracy liability, which properly stated the law. Granted, there might have been more focus on the exact language if defense counsel had been focused on his client's dealings with Earl, but even now Wells doesn't so much as hint at any language he would want altered.

Finally, Wells argues he would have conducted cross examination differently. Again he gives no specifics. Moreover, his trial defense was a general denial, meaning that he had as much reason to counter all testimony of his drug-selling activity as he would have had if the indictment had named only him and Earl. (We note that the same goes for Baugham, who adopted a "lone wolf" defense, meaning he had as much reason to refute any of Earl's testimony that suggested a conspiratorial relationship as he would have had if the indictment had named only him and Earl.) Even under the government's dominant theory at trial, both appellants had ample incentive to attack the government's evidence as to their drug-dealing links with each of the charged co-conspirators.

Assuming the government bears the burden of persuasion in harmless error cases involving variance (as it surely does in harmless error cases generally), the government has successfully carried that burden here.

Sentencing the appellants before the Supreme Court's decision in United States v. Booker, 543 U.S. 220, 125 S. Ct. 738, 160 L. Ed. 2d 621 (2005), the district court treated the U.S. Sentencing Guidelines as mandatory. Under Booker, of course, this violated their Sixth Amendment rights. Though the government's brief conceded that at trial the appellants had preserved their claims under the principle established in Booker, it erroneously suggested that such a preserved claim merited only a limited remand under United States v. Coles, 403 F.3d 764 (D.C.Cir.2005), to seek the district court's position on whether it would have imposed a different sentence had it recognized its discretion. Government counsel at oral argument corrected this mistake and conceded that the proper remedy in this case is a remand for full-blown resentencing. So the sentences will be vacated and the case remanded.

Appellants do, however, make one sentencing claim that we may resolve. They invoke United States v. Brisbane, 367 F.3d 910 (D.C.Cir.2004), which, as we noted at the outset of the opinion, holds that convictions dependent on "cocaine base" under 21 U.S.C. § 841 require evidence that the cocaine at issue was either crack or smokable (leaving unresolved whether proof of smokability alone was enough). Id. at 914. Appellants argue that the evidence against them fails to establish these characteristics. In addressing appellants' sufficiency and variance attacks on the conspiracy conviction we considered Brisbane on our own initiative, finding evidence supporting each appellant's participation in a vertical conspiracy involving far more than the 50 grams of crack for which each was convicted of conspiring to distribute.

Appellants did not make a Brisbane attack on their substantive convictions below, so we review for plain error under Federal Rule of Criminal Procedure 52(b). Under the familiar Olano criteria, an affected conviction must be reversed if (1) there is error (2) that is plain and (3) that affects substantial rights, and (4) we find that the error "seriously affects the fairness, integrity or public reputation of judicial proceedings." Olano, 507 U.S. at 732, 113 S. Ct. 1770 (citations, internal quotation marks, and brackets omitted). The Supreme Court has left unresolved whether, where the law on a point was simply nonexistent at the time of trial (as it was here, since Brisbane was decided after appellants were sentenced and pre-Brisbane cases contained no strong signs one way or the other), an error under the law at the time of appeal can be regarded as plain. Johnson v. United States, 520 U.S. 461, 467-68, 117 S. Ct. 1544, 137 L. Ed. 2d 718 (1997); see also United States v. Johnson, 437 F.3d 69, 74 (D.C.Cir.2006). As in our Johnson decision, we assume in appellants' favor that, if there was any error, it was plain. Thus we turn to the third element of Olano, the requirement that appellants "make a specific showing of prejudice," 507 U.S. at 735, 113 S. Ct. 1770, i.e., show that the error "affected the outcome of the district court proceedings," id. at 734, 113 S. Ct. 1770. Appellants fail to make that showing.

Wells confines his discussion of the record to substantive count 23 in the indictment against him, involving the cocaine that Perry Nelson purchased from him in January 2000. He contends that the evidence was insufficient to show that it was crack. We disagree. Witkowski, who personally received the cocaine from Perry, testified that it was a "white rock substance." Witkowski made clear that he distinguished between rock and powder when he testified at another point that he once recovered "powder cocaine" during a separate operation. Further, Witkowski agreed with examining counsel that "cocaine base" was "another name" for "[c]rack" and that crack was the "vernacular slang for cocaine base." Further, witnesses who testified to Wells's regular drug-dealing referred expressly to (or implicitly accepted the prosecutor's statements regarding) reliance on Wells for crack, without express reference to any other drug. For his part, Wells does not suggest that he dealt a non-crack form of cocaine.

In some ways, the foregoing evidence is less suggestive of crack than in Johnson, but in others it is more so. In Johnson, the government pointed to evidence that the cocaine's purity level was comparable to that of crack, that its packaging was typical of crack, and that defendant had a "cooking kit" (though not shown to be useful only for producing crack), Johnson, 437 F.3d at 75; here, those items are missing. But there are offsetting strengths to the government's case—Wells's notoriety as a crack dealer, the detective's understanding that "cocaine base" is generally synonymous with crack, and the detective's ability to distinguish between rock and powder. And, as in Johnson, 437 F.3d at 75, an officer described the substance as a "white rock." Overall, Wells has failed to carry his burden of showing that the evidence was inadequate.

For his part, Baugham gives no citations to the record but simply makes a series of general assertions to the effect that the witnesses spoke only of cocaine base. These blanket assertions do not hold up. For example, Witkowski testified that the cocaine seized from Young after he received it from Baugham was "crack cocaine" and that the cocaine seized from Baugham's car in numerous ziploc bags was "a white rock substance." The record is, at the very least, more complex than Baugham's blanket assertions suggest, and without more specifics, he cannot carry his burden of showing prejudice.

* * * * * *

Appellants' convictions are affirmed. Their sentences are vacated and their cases are remanded for resentencing not inconsistent with this opinion.

So ordered.