Section 4A1.3(e) provides that "if reliable information indicates that the criminal history category does not adequately reflect the seriousness of the defendant's past criminal conduct or the likelihood that the defendant will commit other crimes, the court may consider imposing a sentence departing from the otherwise applicable guideline range. Such information may include, but is not limited to, information concerning ... prior similar adult criminal conduct not resulting in a criminal conviction." We have italicized the word on which the first issue pivots. For while the guideline is explicit that information justifying an adjustment in the criminal history category is not limited to conduct that falls within one or more of the specific categories in the guideline, such as prior similar adult criminal conduct not resulting in a criminal conviction, the judge based the adjustment in Anderson's criminal history solely on the similarity, as the judge saw it, of the rapes and false imprisonment to the offense of conviction, the drug offense. The text of the guideline and our interpretation of that text make clear that the judge could have based his upward departure on prior dissimilar criminal conduct of Anderson, United States v. Schweihs, 971 F.2d 1302, 1319 (7th Cir.1992), but that would be a discretionary judgment on his part and we cannot exercise the discretion that the guidelines lodge in him. We note that the government does not argue that the sentence should be upheld on an alternative basis.

The record reveals no similarity between Anderson's prior crimes of rape and false imprisonment and the offense of conviction as a lay person would use the term "similar," except, of course, that all are crimes. The drug offense, the offense of conviction, was simple dealing, with no hint of violence; it was the conduct of a business activity that happens to be illegal. The rapes were--rapes. They had nothing to do with Anderson's drug dealing. Neither rape resulted in a prosecution, though the evidence presented at the sentencing hearing that Anderson had indeed committed them was adequate. The false imprisonment did result in a prosecution, and indeed a conviction, but for a different crime. Anderson abducted two men who, he thought, knew the whereabouts of another person. He drove them around in his car for several hours, beat them, and fired a bullet that grazed the side of one of them. Anderson pleaded guilty to a misdemeanor battery of one of the men, when the two victims declined to press charges. The evidence presented at the sentencing hearing that his crime was indeed false imprisonment rather than mere battery was, again, adequate. The abduction had nothing to do with Anderson's drug dealings, however; it was his response to threats of retaliation that had been made against him because of a previous rape that he had committed.

The government argues that all violent crimes are similar to one another. Even if this is right, as we doubt, its application to this case is unclear. Not all drug offenses are violent; Anderson's was not. But, ripostes the government, they are treated like violent offenses even when they are nonviolent; and the government cites statutory and guidelines provisions in which the same treatment is meted out to a defendant convicted or accused of a violent crime or a felony drug offense. E.g., 18 U.S.C. Secs. 924(e), 3142(f)(1); U.S.S.G. Sec. 4B1.1. These and other laws group felony drug offenses with violent crimes for a number of purposes, not only sentencing but also pretrial detention and even deportability. They undoubtedly reflect a legislative determination that felony drug offenses are as serious in some broad sense as crimes of violence, though the most serious crimes of violence, such as murder, are punished more heavily than the least serious felony drug offenses. But not all serious crimes are similar to each other. There is no trick in finding "paper" offenses, as the government calls them, that are punished as severely as crimes of violence. An example is counterfeiting, the maximum penalty for which--20 years in prison--exceeds that for many felony drug offenses. Compare 18 U.S.C. Sec. 470 with 21 U.S.C. Sec. 841(b)(1)(D). Yet the government does not consider counterfeiting similar to a felony drug offense.

The government's position implies, paradoxically, that if a minor drug dealer commits a rape for which he is not convicted, the rape, being "similar" conduct, can be used under section 4A1.3(e) to increase his criminal history score and hence his punishment, while if a counterfeiter, though a more serious criminal in the eyes of Congress, commits a rape, his rape cannot be so used. The government's only suggestion as to how this pattern might be thought rational is that drug offenses are on the whole more serious than "paper" offenses even though some of the latter are more serious than some of the former. Maybe so, but we do not see what this has to do with similarity.

We asked the government's lawyer at argument what penological purpose lies behind the limitation of section 4A1.3(e) to similar crimes, and we got no answer. It is difficult to apply a statute or rule without a clue to its purpose or function, and one hesitates to do so blindly when the consequence is to add years to a person's prison sentence. The guidelines do not furnish a clue to the purpose of "similar." In the absence of guidance there is little alternative to taking the word in its ordinary-language sense, and then the government loses.

Here is a stab at the purpose or function of the provision. It is a provision, remember, for adjusting a defendant's criminal history score when his record of actual convictions understates his criminal history. The focus will therefore be on crimes that the defendant committed but was not convicted of or with respect to which he was allowed to plead to a lesser offense, as in the case of the false imprisonment committed by Anderson. If these prior crimes are similar to the offense of conviction, the inference arises that the defendant through aptitude or experience has become skilled in the commission of a particular type or class of crimes and as a result is able to get away with most of them. This would be a clear ground for a heavier sentence. It was a factor present (though not emphasized) in United States v. Dzielinski, 914 F.2d 98, 101 (7th Cir.1990), where we upheld (though citing subsection (a) rather than (c) of section 4A1.3 because the defendant had been convicted of the previous crimes) a finding that the defendant's earlier crimes of fraud and forgery were similar to the offenses of conviction, which were bank robberies. The defendant acknowledged having robbed the banks in order to pay off the victim of one of his frauds.

If we are right about the purpose of the guideline--and we derive some support for our speculation from United States v. Schmude, 901 F.2d 555, 559 (7th Cir.1990), and United States v. Molina, 952 F.2d 514, 519 (D.C.Cir.1992)--it suggests that offenses as unlike each other in motive and method as drug dealing and rape are not similar within the meaning of the provision. The false imprisonment presents a more difficult question. There is a hint in the record that the reason the victims of the abduction did not press charges is that they were intimidated by the sight of Anderson and his associates at the police station. If the associates were Anderson's partners in the drug business, this may be a case where the existence of a criminal gang formed for one purpose facilitates the members' commission of their "personal" crimes, as in United States v. Masters, 924 F.2d 1362, 1366-67 (7th Cir.1991). Dissimilar past crimes might be thought similar to the offense of conviction if facilitated by the activities giving rise to that offense, or, as in Masters, if the dissimilar past crimes facilitated the offense of conviction.

We do not wish to speak dogmatically on this difficult question--plainly one that the Sentencing Commission ought to address. The district court's analysis was inadequate and the case must therefore be returned to it for resentencing.

The second issue relates to the weight that the judge gave the false imprisonment in his recomputation of Anderson's criminal history score. The issue will arise on remand if the judge either finds that the false imprisonment was similar conduct to the offense of conviction on some basis that we have not thought of or decides to base an upward adjustment on dissimilar conduct. Anderson had, we recall, been convicted of misdemeanor battery for what was actually a false imprisonment. If the conviction had been for false imprisonment and if upon conviction for false imprisonment Anderson had received a prison sentence of at least 13 months, then he would have received a higher criminal history score than he received for his conviction of battery. Section 4A1.1(a) of the guidelines awards three criminal history points for a sentence of imprisonment that exceeds a year and a month, and Anderson had received only two points for the misdemeanor battery conviction. The evidence at the sentencing hearing showed that Anderson had committed false imprisonment, and the district judge, on the basis of his experience as a former judge of the state court in which Anderson would have been sentenced, was sure that Anderson would have received a sentence of at least 13 months had he been properly charged, that is, charged with false imprisonment.

With the judge's decision to reclassify the battery as false imprisonment we have no quarrel, provided--the issue for further exploration on remand--the reclassified offense, for which Anderson was not convicted, can be fitted within section 4A1.3 of the guidelines. But we do not think it was within the judge's authority to speculate about the sentence that Anderson would have received had he been convicted of false imprisonment. We have no doubt that it was a well-founded speculation, given the judge's previous experience as a Wisconsin circuit judge. But the guidelines' goal of promoting uniformity in sentencing would be impaired if judges used their personal experience to recompute criminal history scores. A Wisconsin judge sentencing Anderson for false imprisonment would have had complete sentencing discretion within the statutory limits and thus might have sentenced Anderson to a term of prison shorter than 13 months. See Wis.Stat.Ann. Secs. 939.50(3)(e), 940.30. His plea of guilty to battery resulted in a sentence of probation. Who knows what sentence a plea of guilty to a somewhat more serious offense, though involving the same facts, would have yielded?

The choice of sentence under the guidelines is required to be based on considerations that are objective and reliable. U.S.S.G. Sec. 6A1.3(a) and Commentary thereto; United States v. Vold, 66 F.3d 915, 918 (7th Cir.1995); United States v. Beler, 20 F.3d 1428, 1432 (7th Cir.1994). Speculation as to the exercise of discretion by a judge in another jurisdiction concerning hypothetical charges is not in our view objective or reliable. In United States v. Ruffin, 997 F.2d 343, 344 (7th Cir.1993), we disapproved the judge's basing an upward departure on his experience as a prosecutor. The principle here is the same. The district judge should not have used his previous experience to guess at the sentence that Anderson would have received had he been convicted of a crime with which he was not charged.

REVERSED AND REMANDED.

MANION, Circuit Judge, concurring.

I accept the court's decision to remand for resentencing so that the district court can clarify how Anderson's drug conviction is similar to his prior crimes, or perhaps why his prior crimes are dissimilar yet reflective of a more serious criminal history that may deserve a higher sentence.

I write separately to emphasize that "simple" drug dealing in transactions that have no outward sign of violence is not just a "business activity that happens to be illegal" (ante at 3). Illegal drugs such as cocaine, the product Anderson chose to sell, are dangerous, addictive, and destructive to the user. Sellers viciously fight over customers and selling territory. We have consistently held that guns are typical tools of this dangerous trade. In short, while "not all drug offenses are violent," the drug trade and culture are inherently violent. Users will rob and maim to obtain money for their next purchase. Street sellers and large quantity distributors alike inflict wanton violence upon competitors and traitors. See United States v. Boyd, 55 F.3d 239, 241 (7th Cir.1995); United States v. Bates, 843 F. Supp. 437, 440 (N.D.Ill.1994). That is why Congress has insisted upon substantial sentences even for illegal drug transactions that have "no hint of violence."

As with this case, courts must take care to sentence in accordance with the directives of the Guidelines. But courts also cannot diminish the serious damage the drug trade continues to inflict on our society.