Convicted of using unauthorized access devices, see 18 U.S.C. &#167 1029(a)(2), (e)(1), Clarence Trotter received a mild sentence: 5 months' imprisonment and 36 months' supervised release (5 months of which were to be spent in home confinement). He repeatedly violated the terms of home confinement by leaving without authorization, and he broke the rules of supervised release by lying to the probation office about his employment (and failing to notify the office about changes) and using marijuana. He tested positive for marijuana at least 3 times and skipped 7 scheduled drug tests. Trotter also paid only $3,800 of the $18,300 restitution that is part of his sentence--and only $80 of that sum voluntarily (the rest came from garnishing his salary). Eventually the district judge revoked his supervised release.

Under the Sentencing Guidelines, the penalty following revocation depends on the most serious violation-- for Trotter, unlawful possession of drugs. Multiple violations are not cumulative under the Guidelines' structure, U.S.S.G. &#167 7B1.1(b), so the district court ignored all of Trotter's other shortcomings. The Sentencing Commission has prescribed three grades of violations--A, B, and C. (Chapter 7 of the Guidelines is a Policy Statement rather than a rule, see United States v. Hill, 48 F.3d 228 (7th Cir. 1995), but a district court must start with these policies, and in Trotter's case the judge did not show any inclination to deviate from them. Thus for current purposes we treat Chapter 7 as if it were binding.)

A crime of violence, a firearms offense, or a "controlled substance offense" is a Grade A violation; any other conduct "constituting any other federal, state, or local offense punishable by a term of imprisonment exceeding one year" is a Grade B offense (U.S.S.G. &#167 7B1.1(a)(2)); any remaining infraction falls into Grade C. Simple possession and use of a drug, without intent to distribute, is not a "controlled substance offense" under the chain of cross-references starting with Application Note 3 to &#167 7B1.1 and leading to U.S.S.G. &#167 4B1.2(b) and its Application Note 1. Use of marijuana is not punishable by more than a year in prison under either state or federal law, making it a Grade C violation. Possession of personal-use quantities is a civil offense punishable by a fine, 21 U.S.C. &#167 844a, but this dispensation is not available to persons with a drug conviction, see &#167 844a(c). Trotter has such a conviction, which moves him to the domain of 21 U.S.C. &#167 844(a): Simple possession by a repeat offender is punishable by up to two years' imprisonment. This led the district judge to conclude that Trotter committed a Grade B violation. Because Trotter's criminal history category is II, the table at &#167 7B1.4 gives a range of 6-12 months' imprisonment. The judge selected 6 months, plus an additional 30 months' supervised release. The range for a Grade C violation is 4-10 months, so Trotter could have received the same sentence no matter what. But the judge did not state on the record that the classification is irrelevant, and we must therefore tackle the merits, see United States v. Mount, 966 F.2d 262 (7th Cir. 1992); perhaps the judge would have chosen 4 months had he believed that Trotter's misconduct is Grade C rather than Grade B.

This potential 2-month difference in a term of imprisonment has spawned a cascade of legal issues that reveal disagreements within the circuit. No fewer than three intra-circuit conflicts have come to light. The first is whether expiration of the time in prison moots the choice between Grade B and Grade C. If the case is not moot, we must decide whether use of marijuana permits the district judge to infer that the user committed the crime of possessing that drug. If Trotter possessed marijuana, the final question is whether the district judge could consider his criminal history, which elevated his potential sentence from a $10,000 fine to two years' imprisonment, and thus from Grade C to Grade B. We start with mootness.

Trotter's imprisonment is over, but he remains on supervised release, a form of custody. If instead of being sentenced to 6 months Trotter had been sentenced to only 4, what would have happened to the length of the supervised release? If time removed from imprisonment were added automatically to supervised release, then nothing we do now could assist Trotter, and it might make things worse by tacking months onto the remaining term of custody. But this is not how either the supervised-release statute, 18 U.S.C. &#167 3583, or the Guidelines, handles matters. So far as the statute is concerned, a combination of prison and supervised release adding to 36 months is the maximum available to the judge, see &#167 3583(h), but not the minimum. The Guidelines do not detract from that statutory flexibility. See U.S.S.G. &#167 7B1.3(g)(2). The district judge could have sentenced Trotter to 4 months' imprisonment and, say, 24 months' supervised release--and maybe he would have done so had he believed that Trotter's deeds were Grade C rather than Grade B violations. We cannot be sure, of course, but Trotter is entitled to have the district judge impose a sentence under the correct legal rules, if any potential benefit could arise from the difference. This principle is the reason why, as Mount held, we consider his claims even though 6 months' imprisonment could have been meted out for a Grade C violation. Unless we are confident that Trotter cannot benefit from success on appeal, the case is not moot.

That is how things work out from the application of first principles. It is also how this court saw matters in United States v. Swigert, 18 F.3d 443 (7th Cir. 1994), and United States v. Eske, 925 F.2d 205, 206 n.2 (7th Cir. 1991). See also United States v. Verdin, 243 F.3d 1174, 1177-79 (9th Cir. 2001). Swigert relied in part on a proposition later rejected in United States v. Johnson, 529 U.S. 53 (2000): That a person held too long in prison automatically would be given credit against his term of supervised release. But Swigert does not depend wholly on that proposition; it also recognized that on remand a district judge would have discretion to shorten the term of supervised release.

United States v. Ross, 77 F.3d 1525, 1549 n.6 (7th Cir. 1996), throws a monkey wrench into this approach by holding that completion of a prison sentence while an appeal is pending does moot an appeal, despite ongoing supervised release. Ross recognized the seemingly contrary holdings in Swigert and Eske but added that the appeal is moot when it is legally impossible to shorten the term of supervised release, even if the prison sentence should have been shorter. That unexceptional proposition of law seems to entail the proposition that it is legally forbidden to shorten the period of supervised release. Sometimes that might be so; minimum terms accompany some offenses, and after Johnson a defendant who received the minimum term of supervised release has nothing further to gain with respect to supervised release from a post-hoc declaration that his imprisonment should have ended sooner. But in Ross itself the district judge had discretion to select a term of supervised release as low as 24 months, which would have meant a reduction from the term actually imposed, 36 months, just as in Swigert and Eske. (The appellant had been convicted of mail fraud and related Class D offenses, which is why the district judge possessed discretion to choose a reduced term of supervised release. See 18 U.S.C. &#167 &#167 152, 3559(a); U.S.S.G. &#167 5D1.2(a).) Ross therefore might be understood to create an intra-circuit conflict. We can avoid that, however, by limiting Ross to the proposition it said was dispositive: if it is impossible to reduce the term of supervised release, a defendant who is already out of prison has nothing to gain from an appeal. Ross did not analyze when it is possible to reduce that term; it made an assumption on that subject, but assumptions are not holdings. Ross therefore does not preclude us from deciding that Trotter's case is live.

Next comes the question whether use of marijuana supports an inference of possession--for only possession is a criminal offense potentially supporting a Grade B violation. This question has a simple answer. Of course use may support an inference of possession. A person might test positive for marijuana if he never touched the stuff but spent a lot of time around heavy smokers and ended up with some of the drug in his lungs. This is unlikely, but possible, and a district judge therefore might treat one or even two positive tests as inadequate to prove possession of the drug. But a judge also could infer from the presence of marijuana metabolites in the blood that the user possessed that drug, if only in the process of smoking a communal joint. Trotter tested positive at least 3 times and skipped a further 7 tests, which could support a conclusion that he knew that he would have tested positive on those occasions too. That is more than enough to permit a rational trier of fact to infer that Trotter possessed marijuana. The burden of persuasion for revocation of supervised release is a preponderance of the evidence, and the finding of possession on a record such as this cannot be deemed clearly erroneous. Inferring possession of a drug from the consumption of that drug is just as sensible as inferring, from the statement "I ate a hamburger for lunch," that the person possessed the hamburger before wolfing it down. Likewise judges and juries may infer that someone who robbed a bank also possessed the loot. No surprise, then, that we held in United States v. Young, 41 F.3d 1184, 1185 (7th Cir. 1994), that use of a drug implies its possession. (Indeed, Young suggested that use compels an inference of possession, a suggestion that may be overboard for a single positive drug test.) Every other court that has considered this question has come to the same conclusion. See, e.g., United States v. Dow, 990 F.2d 22, 24 (1st Cir. 1993); United States v. Blackston, 940 F.2d 877, 878 (3d Cir. 1991); United States v. Clark, 30 F.3d 23, 25 (4th Cir. 1994); United States v. Courtney, 979 F.2d 45, 49 (5th Cir. 1992); United States v. Crace, 207 F.3d 833, 835-37 (6th Cir. 2000); United States v. Baclaan, 948 F.2d 628, 630 (9th Cir. 1991); United States v. Rockwell, 984 F.2d 1112, 1114 (10th Cir. 1993); United States v. Almand, 992 F.2d 316, 318 (11th Cir. 1993).

As we said, the question whether use of a drug may support an inference of its possession has a simple answer-- "yes," according to Young and eight other circuits. Unfortunately, a different panel of this court gave the answer "no" in United States v. Wright, 92 F.3d 502 (7th Cir. 1996). Without citing Young or the decisions in other circuits, the panel in Wright opined that "testing positive for drug use is not an independent crime; it is only a violation of the terms of . . . supervised release. We note . . . that these kinds of violations are classified as Grade C under &#167 7B1.1(a)(3)." 92 F.3d at 506. Wright may have meant to say no more than that if the district judge declines to infer possession from use, then no "independent crime" has been committed. But the flat assertion (which appears twice in Wright) that persons who test positive for drug use "are classified as Grade C" violators implies that it would be improper to infer possession from use. That position is impossible to sustain. It would take normal fact-finding and inference-drawing out of the hands of district judges. Wright is best understood not to do such a thing. The statement that persons who test positive for drug use "are classified as Grade C" violators is dictum in Wright--the district judge had found a Grade A violation, and the choice between Grade B and Grade C was not presented by the case. Unreasoned statements in our decisions should not be taken to overrule opinions of prior panels, such as Young. We disapprove the dictum in Wright and reiterate the holding of Young, which has such substantial support in other circuits (plus the norm committing inferences to the trier of fact).

Finally, we must decide whether Trotter's possession of marijuana was conduct "constituting any other federal, state, or local offense punishable by a term of imprisonment exceeding one year" within the meaning of &#167 7B1.1(a)(2). The answer is yes if the district court may consider Trotter's prior drug conviction, and no otherwise. "Yes" seems an apt answer; how can one determine whether an offense is "punishable" by a particular term without considering all of the ingredients that set the maximum punishment? If a statute punished marijuana possession by two years in prison if the possession occurred within 1,000 feet of a school, and one month otherwise, the court would have to determine the distance from the school to know whether given conduct was punishable by more than a year in prison. Just so with a recidivist enhancement. See Almendarez-Torres v. United States, 523 U.S. 224 (1998). Many provisions of the Guidelines other than &#167 7B1.1(a)(2) make something turn on whether conduct is punishable as a felony, by a sentence exceeding one year, or something similar. We routinely hold that when making these decisions the court must determine whether the conduct is a felony (etc.) after prior convictions are taken into account. See, e.g., United States v. Irby, 240 F.3d 597, 599-600 (7th Cir. 2001); United States v. Rice, 116 F.3d 267, 268 (7th Cir. 1997). Young reached the same conclusion for purposes of &#167 7B1.1(a)(2), although without spelling out the reasoning. 41 F.3d at 1186 & n.1. And both other courts of appeals that have addressed the question are in accord with Young. See Crace, 207 F.3d at 837-38; United States v. Boisjolie, 74 F.3d 1115, 1116-17 (11th Cir. 1996).

Young seems to be the Rodney Dangerfield of this circuit's opinions. Wright contradicted one of Young's holdings without mentioning it, and in United States v. Lee, 78 F.3d 1236 (7th Cir. 1996), we contradicted Young's other holding, again without recognizing the state of circuit law. One reason why this aspect of Young hasn't received much respect is the indirect way in which the panel in Young stated the point, in a cursory footnote. Apparently the lawyers in Lee did not draw that panel's attention to Young. But it has been drawn to our attention, and we must decide what to do. One case or the other must be overruled.

Lee gave two principal reasons for holding that prior convictions should be disregarded when determining the grade of a violation: first that Application Note 1 to &#167 7B1.1 supports this approach, and second that because the table in &#167 7B1.4 includes a criminal history category, use of criminal history to influence the grade level would be double counting.

Under 18 U.S.C. &#167 &#167 3563(a)(1) and 3583(d), a mandatory condition of probation and supervised release is that the defendant not commit another federal, state, or local crime. A violation of this condition may be charged whether or not the defendant has been the subject of a separate federal, state, or local prosecution for such conduct. The grade of violation does not depend upon the conduct that is the subject of criminal charges or of which the defendant is convicted in a criminal proceeding. Rather, the grade of the violation is to be based on the defendant's actual conduct.

Lee relies on the last sentence: "the grade of the violation is to be based on the defendant's actual conduct." This means, we thought, that the court should put to one side earlier offenses, which are not part of the "actual conduct" that occurs while on supervised release. Putting this sentence back into context, however, gives it a different meaning. Application Note 1 tells the district judge to consider what the person on supervised release did, rather than what crimes he has been charged with. Revocation of supervised release, in other words, proceeds on real-offense rather than charge-offense principles. A judge engaged in real-offense sentencing does not ignore prior offenses that affect the maximum punishment; recidivist enhancements are part of real-offense sentencing.

As for double counting: If the Guidelines' use of criminal history to influence the final sentence blocks recidivist enhancements under &#167 7B1.1, it logically does so for any offense and not just for revocation of supervised release, for the use of criminal history is ubiquitous in the Guidelines. Yet this has never before been thought incompatible with using prior convictions to determine maximum permissible punishments. Just one year after we released Lee, the Supreme Court rejected this element of its reasoning in United States v. LaBonte, 520 U.S. 751 (1997). Career criminals must be sentenced at or near the maximum term for their crimes. 28 U.S.C. &#167 994(h). Does "maximum term" for this purpose mean the maximum with or without recidivist enhancements? The first circuit concluded in LaBonte that the maximum must be determined without recidivist enhancements, because the effect of prior convictions is taken into account in the career-criminal enhancement (as well as the criminal history category system). But the Supreme Court held otherwise, concluding that the term of punishment to which a person is exposed on violating a statute includes all enhancements--for quantity of drugs, for use of firearms, for violence during the offense, and for prior convictions. A similar approach applied to &#167 7B1.1(a)(2) supports Young rather than Lee, which must therefore be overruled. This step will restore harmony within the circuit and eliminate a conflict among the circuits.&#134

Trotter's prior drug conviction exposed him to a maximum of two years in prison for possessing marijuana. The district judge did not commit a clear error in concluding that Trotter possessed marijuana in the course of using it. Trotter therefore committed a Grade B violation, and the punishment imposed on the revocation of his supervised release is

Affirmed.