LOKEN, Chief Judge.

Donald Lee Houston pleaded guilty to manufacturing, attempting to manufacture, and aiding and abetting the manufacture of five or more grams of actual methamphetamine in violation of 21 U.S.C. §§ 841(a)(1) and 841(b)(1)(B). Houston now appeals his sentence, arguing the district court clearly erred in finding that he is accountable for more than 50 but less than 150 grams of actual methamphetamine, which resulted in a base offense level of 32. The issue on appeal, one of first impression, is whether the government proved that the methamphetamine quantities Houston admitted he helped manufacture were "actual" methamphetamine quantities, rather than mixture quantities. In reviewing the drug quantity finding for clear error, we must affirm unless the entire record firmly convinces us that a mistake has been made. See United States v. Moore, 212 F.3d 441, 446 (8th Cir. 2000). Concluding that the issue was properly preserved in Houston's objections to the Presentence Investigation Report (PSR), and entirely ignored in the government's proof at sentencing, we reverse.

I. Framing the Issue

For drug manufacturing offenses, the Sentencing Guidelines provide that a defendant's base offense level is based upon drug quantity. U.S.S.G. § 2D1.1(a)(3). The Drug Quantity Table in § 2D1.1(c) specifies the base offense levels for various quantities of the most common illegal drugs. For only three drugs — PCP, methamphetamine, and amphetamine — the Table lists alternative quantities for each offense level. For example, the Table places the following quantities in Level 32:

At least 500 G but less than 1.5 KG of Methamphetamine, or at least 50 G but less than 150 G of Methamphetamine (actual). . . .

§ 2D1.1(c)(4). The commentary to § 2D1.1 then explains:

The term . . . "Methamphetamine (actual)" refer[s] to the weight of the controlled substance, itself, contained in the mixture or substance. For example, a mixture weighing 10 grams containing [methamphetamine] at 50% purity contains 5 grams of [methamphetamine] (actual). In the case of a mixture or substance containing . . . methamphetamine, use the offense level determined by the entire weight of the mixture or substance, or the offense level determined by the weight of the . . . methamphetamine (actual), whichever is greater.

§ 2D1.1, comment. n. (B) to Drug Quantity Table (emphasis added). As the above-quoted Level 32 entry illustrates, the alternative quantities for each base offense level in the Table comprise a 10:1 ratio, the same ratio codified by Congress in the mandatory minimum penalty provisions applicable to methamphetamine offenses. See, e.g., 21 U.S.C. § 841(b)(1)(A)(viii).

The government bears the burden of proving drug quantity by a preponderance of the evidence. See United States v. Hollingsworth, 298 F.3d 700, 703 (8th Cir. 2002), cert. denied, ___ U.S. ___, 123 S.Ct. 1307, 154 L.Ed.2d 1061 (2003). When use of the actual methamphetamine alternative produces a greater sentence, the government must prove the actual methamphetamine content of the substance or substances in question. See United States v. Bogusz, 43 F.3d 82, 87 (3d Cir. 1994). The government may prove the total quantity of actual methamphetamine in a series of transactions by testing the purity of a seized quantity and applying the percentage of actual methamphetamine in the tested quantity to the unrecovered quantities. See United States v. Newton, 31 F.3d 611, 614 (8th Cir. 1994); accord United States v. Lopes-Montes, 165 F.3d 730, 731-32 (9th Cir. 1999). When no quantity has been recovered, the government may prove the purity of quantities attributed to the defendant by circumstantial evidence, for example, a conspirator's reliable testimony that purchased methamphetamine was "undiluted, unadulterated . . . not cut . . . pure," United States v. Cockerill, 217 F.3d 841, 2000 WL 852608 at *1 (4th Cir. 2000) (unpublished per curiam), or an expert's testimony as to the normal purity of methamphetamine produced in a lab, see United States v. Mosby, 177 F.3d 1067, 1071 (8th Cir. 1999).

As with all fact-intensive sentencing issues, if the defendant makes a timely objection to the drug quantity determination in the PSR, "if the sentencing court chooses to make a finding with respect to the disputed facts, it must do so on the basis of evidence, and not the presentence report." United States v. Hudson, 129 F.3d 994, 995 (8th Cir. 1997) (quotation omitted).

II. The Sentencing Record in This Case

The determination of drug quantity in this case began, as it usually does, with the factual recitations in the PSR. The PSR set forth two alternative ways to calculate the total quantity of methamphetamine attributable to Houston. First, paragraph 32 of the PSR assessed "a total of 212.6 grams of methamphetamine (actual)." This total quantity was based upon (i) Houston's admission that he had manufactured "½ to ¾ ounce of actual methamphetamine" with co-defendant Chad Curley on five or six occasions, resulting in 70.85 grams of actual methamphetamine (PSR paragraph 28); and (ii) Houston's admission that he allowed Kevin Demoss to manufacture one ounce of methamphetamine on Houston's property on four or five occasions, resulting in an additional 141.75 grams of actual methamphetamine (PSR paragraph 31). Based upon this 212.6 gram finding, the PSR recommended a base offense level of 34 (more than 150 but less than 500 grams of actual methamphetamine).

In his PSR objections, Houston denied the quantities alleged in paragraph 28 and alleged that he had "nothing to do with" the activities of Demoss alleged in paragraph 31. In addition, Houston objected to the assessment of 212.6 grams of actual methamphetamine in paragraph 32 of the PSR, stating that the correct total quantity is 56 grams and "the Base Offense Level should not be 34 but should be Level 26." The Drug Quantity Table places a total quantity of 56 grams of methamphetamine mixture in Level 26. See § 2D1.1(c)(7) ("[a]t least 50 G but less than 200 G of Methamphetamine, or at least 5 G but less than 20 G of Methamphetamine (actual))." Thus, Houston's objections to the PSR clearly raised the actual methamphetamine issue. The sentencing hearing transcript confirms that the district court understood this was a contested issue.

Alternatively, paragraph 32 of the PSR calculated total drug quantity by assessing 814.35 grams of methamphetamine mixture. This quantity was based primarily upon "laboratory reports" that officers recovered 671 grams in a warrant search of Houston's home in November 2001. This alternative quantity produced a base offense level of 32. In his objections, Houston denied "any involvement with the substance" and denied "that this mixture could have made methamphetamine." At sentencing, the government offered no proof regarding what was recovered in the search of Houston's home. Therefore, we may not uphold the district court's base offense level finding by this alternative method.

At the sentencing hearing, the government relied upon the testimony of Iowa Division of Narcotics Enforcement Officer Gregory Brugman to prove the actual methamphetamine drug quantity set forth in the PSR. Agent Brugman testified:

Q. Okay. What did the defendant tell you about manufacturing methamphetamine?

A. He stated — we initially talked to him about Chad Curley and he stated he had helped Mr. Curley on approximately five to six occasions when he manufactured methamphetamine.

* * * * * *

Q. Did he tell you what quantity or approximately what quantity of methamphetamine they cooked each time?

A. He stated he was cooking approximately a half ounce to three-quarters of an ounce quantity of methamphetamine on each occasion.

Later in his direct testimony, Agent Brugman was asked about Houston's admissions regarding Kevin Demoss:

Q. What did he tell you about that date?

A. He stated that Chad Curley had come out to the farm along with his cousin, Keven Demoss, and at that time they had done a one ounce cook of methamphetamine.

* * * * * *

Q. Did he give you any more information about this Kevin Demoss and cooking methamphetamine?

A. Yes, he did.

Q. What did he say about him?

A. He stated that Kevin Demoss and his wife Dawn lived in his farmhouse with [Houston] and his wife and they were there approximately two or three months, and during that time he witnessed them cook methamphetamine at his residence.

* * * * * *

Q. Did the defendant receive anything for allowing those [cooks] to happen on his property?

A. Yes, he did.

Q. What did he receive?

A. He was getting anywhere from two grams of methamphetamine up to an eighth ounce of methamphetamine.

The district court credited Agent Brugman's testimony in finding that Houston helped Curley produce 70.85 grams of actual methamphetamine. After reducing the quantity attributable to the Demoss cooks that Houston assisted, the court found that Houston was responsible for a total of less than 150 grams of actual methamphetamine, rather than the 212.6 grams calculated in the PSR, and therefore found that his base offense level is 32.

III. Discussion

The district court properly relied on Houston's admissions in estimating drug quantity. The court believed Agent Brugman's testimony regarding those admissions, rather than Houston's vague testimony at the sentencing hearing that his quantity "guesstimates" were overstated. That credibility determination is virtually unreviewable on appeal. See United States v. Behler, 14 F.3d 1264, 1273 (8th Cir.), cert. denied, 513 U.S. 960,

115 S.Ct. 419, 130 L.Ed.2d 335 (1994). Thus, the district court's quantity finding is well-supported. The issue is whether the government proved that the quantities to which Houston admitted were actual methamphetamine, rather than a mixture.

In reviewing this issue, we conclude it is critical that Agent Brugman testified that Houston admitted to helping manufacture methamphetamine, not actual methamphetamine. At no point did government counsel ask Brugman, an experienced drug investigator, whether Houston ever referred to actual or pure methamphetamine. Nor did counsel ask Brugman whether he understood Houston to be referring to quantities of actual methamphetamine. Likewise, when Houston testified for the defense and asserted that the methamphetamine he helped Curley manufacturer was low quality ("filth"), government counsel did not ask on cross examination whether the quantity estimates he gave Agent Brugman referred to the total weights of the substances manufactured by Curley and Demoss, or to Houston's estimate of the net weight of the methamphetamine contained in those substances.

Agent Brugman's testimony was the government's entire case at sentencing on the issue of actual methamphetamine quantity. None of the methamphetamine manufactured by Curley and Demoss was recovered and tested for purity. No expert testified as to the typical purity of methamphetamine manufactured in these quantities, by this method, and in these conditions. Agent Brugman admitted that two samples attributed to Houston — not necessarily end products of the Curley and Demoss cooks — had been tested and found to be methamphetamine mixtures. (The lab reports of these tests were not made part of the record on appeal, and Agent Brugman was not asked what purity data was reported.)

On this record, we can only conclude that the district court, like the author of the PSR, simply assumed that the quantities Houston admitted to Agent Brugman were actual methamphetamine quantities. We think the assumption unwarranted. When a lay person is asked in general terms how much methamphetamine he helped someone else cook, his answer will almost certainly be in terms of the size of the resulting mixture, not the net weight of one of its components. Government counsel probably knew that; in any event, he did not address this critical issue in questioning Agent Brugman and Houston at the sentencing hearing. The result is a record that has no evidence supporting the quantity finding of more than 50 grams of actual methamphetamine. The quantity itself is supported, but the finding that it was actual methamphetamine is clearly erroneous.

Relying upon Agent Brugman's testimony that methamphetamine cooks attempt to produce pure methamphetamine, the government argues that a base offense level of 32 is warranted even if Houston's admissions reflected total quantities produced by Curley and Demoss, rather than the net weight of the methamphetamine contained in those substances. The government cites no authority for this contention, and we reject it. Note (B) to the Drug Quantity Table defines actual methamphetamine as "the weight of the controlled substance, itself, contained in the mixture." (Emphasis added.) Congress and the Sentencing Commission have mandated alternative methods for determining total quantities of methamphetamine, PCP, and amphetamine. The distinction would be obliterated if the total quantity of every batch of methamphetamine was deemed to be actual methamphetamine because the cook hoped or attempted to produce the drug in its pure form.

The record is devoid of any evidence to justify a finding of at least 50 grams of actual methamphetamine, or 500 grams of a methamphetamine mixture, the minimum alternative quantities necessary to place Houston's offense in base offense level 32. Accordingly, the judgment of the district court is reversed and the case is remanded for resentencing on the existing sentencing record.