dissenting.
Milwaukee’s “knock and talk” process is a distraction. What the police planned or tried to do at Apartment 7 is irrelevant; only what they did matters. California v. Hodari D., 499 U.S. 621, 111 S.Ct. 1547, 113 L.Ed.2d 690 (1991). What they did was approach a suspected crime den — as they were entitled to do, whether or not they had probable cause, see Florida v. Bostick, 501 U.S. 429, 111 S.Ct. 2382, 115 L.Ed.2d 389 (1991); United States v. James, 40 F.3d 850, 856 (7th Cir.1994) — and ask a single person to stand still for a brief time until they determined whether the situation was hazardous. My colleagues hold that police must let someone like Johnson go without knowing whether he is armed, even though within moments he may wheel and fire. Reason to believe that persons situated similarly to the subject commonly carry guns — a proposition true about drug distribution points — is inadequate, the majority concludes. The evidence cannot be systemic; it must be particularistic. The distinction between systemic and particular-istic evidence is elusive; ex ante all suspicion is based on categorical inferences (for example, that bulges in clothing are associated with guns). So to say that X was found in a crack house, and that many denizens of such places are armed, is to know a great deal *722about X, more than the police knew in Terry. Let that pass. The point is that the fourth amendment permits reasonable searches and seizures. A stop is reasonable when the subject is in the company of people who often go armed.
Like the majority, I accept the district judge’s finding that'the police did not see a crack pipe inside Apartment 7. They knew some things about that apartment without seeing inside, however, and could infer more. The landlord suspected that it had become a drug distribution point; a maintenance worker at the building “had observed a lot of people going to apartment 7, leaving quickly, and then departing through the rear building exit.” 170 F.3d 708, 711. These observations imply a retail business of some kind, and the lack of signs announcing the nature of the establishment (plus the fact that the business was in a residential building) meant that the operation almost certainly was unlawful. People do not “quickly” leave New Year’s Eve parties by the back stairs, and the maintenance worker’s report of heavy traffic was not limited to the day the officers arrived. Which law the occupants were breaking could not be known without inspection: perhaps they were selling stolen merchandise, pirate videotapes, or bogus “Rolex” watches; quick departures are not consistent with an unlicensed bar; but the activity was consistent with drug distribution, just as the owner suspected. And when the police arrived they heard enough voices from inside to suggest that their hands would be full if the occupants turned out to be hostile. What the officers had was more than a stray report; they had a reasonable suspicion of ongoing crime.
This was the setting from which Johnson emerged, before the officers knocked on the door. Johnson, obviously not a pizza delivery boy or other innocuous type, attempted to walk past the officers; meanwhile sounds within the apartment conveyed a sense of panic once the officers announced their identity. Johnson was wearing bulky clothes that made it easy to conceal weapons. Officer Reilly asked Johnson whether he was armed; Johnson remained mum. Reilly asked Johnson to step back into the apartment; Johnson refused and tried to leave, flailing his arms to break Reilly’s hold on his sleeve. It took three officers to wrestle Johnson to the ground, and in the process Officer Bishop felt a gun. This turned out to be a loaded .357 Magnum. Johnson also was carrying cocaine packaged for retail distribution, plus a second handgun.
My colleagues approach the case as if the question were whether the police had enough suspicion to conduct a Terry frisk. Posing the question in this way ordains the answer, because Ybarra v. Illinois, 444 U.S. 85, 100 S.Ct. 338, 62 L.Ed.2d 238 (1979), holds that the police may pat down a person for weapons only if they have some suspicion in addition to the person’s presence at the scene of a crime. See also Maryland v. Buie, 494 U.S. 325, 334-35 n. 2, 110 S.Ct. 1093, 108 L.Ed.2d 276 (1990); Michigan v. Summers, 452 U.S. 692, 695 n. 4, 101 S.Ct. 2587, 69 L.Ed.2d 340 (1981); Brown v. Texas, 443 U.S. 47, 50-51, 99 S.Ct. 2637, 61 L.Ed.2d 357 (1979); United States v. Evans, 994 F.2d 317, 322 (7th Cir.1993). But it is the wrong question. The police did not frisk Johnson until after feeling a weapon in his pocket during the struggle to prevent his departure. Perhaps they wanted to frisk him earlier (one officer held out his hands to initiate the process), but they did not do so. The right question therefore is whether the police were entitled to insist that Johnson return to the apartment (or remain in sight in the hall) until the nature of the business being conducted, and the risk the occupants posed, had been ascertained. If the police were entitled to detain Johnson for these few minutes, then his attempt to escape and the discovery of one gun was all the cause needed for everything that ensued. United States v. Sharpe, 470 U.S. 675, 682-83 n. 3, 105 S.Ct. 1568, 84 L.Ed.2d 605 (1985); United States v. Quinn, 83 F.3d 917, 921-22 (7th Cir.1996); United States v. Weaver, 8 F.3d 1240, 1244 (7th Cir.1993); Tom v. Voida, 963 F.2d 952, 957-58 & n. 4 (7th Cir.1992); United States v. Chaidez, 919 F.2d 1193, 1200 (7th Cir.1990); United States v. Brown, 159 F.3d 147 (3d Cir.1998). See also Michigan v. Chesternut, 486 U.S. 567, 576, 108 S.Ct. 1975, 100 L.Ed.2d 565 (1988) (Kennedy, J., concurring) (“respondent’s unprovoked flight gave the police ample cause to stop him.”).
*723Two years after Ybarra, the Supreme Court decided that the requirement of person-specific information required for a search (including a pat-down search) did not apply to a seizure, and in particular to a demand that persons remain at the scene of an investigation. Police about to execute a search warrant encountered someone on the front steps of the house to be searched. They did not know whether this person was dangerous and had no reason to suspect that he was carrying a gun but insisted that he stay put while the search was conducted. The Court held in Michigan v. Summers that such a seizure is proper because of the risk that similarly-situated persons will interfere with the search (or induce someone else to do so) if free to leave, and in the process distinguished between detention and detention-plus-frisk:
The “seizure” issue in this case should not be confused with the “search” issue presented in Ybarra v. Illinois, 444 U.S. 85. In Ybarra the police executing a search warrant for a public tavern detained and searched all of the customers who happened to be present. No question concerning the legitimacy of the detention was raised. Rather, the Court concluded that the search of Ybarra was invalid because the police had no reason to believe he had any special connection with the premises, and the police had no other basis for suspecting that he was armed or in possession of contraband. See id., at 90-93. In this case, only the detention is at issue. The police knew respondent lived in the house, and they did not search him until after they had probable cause to arrest and had done so.
452 U.S. at 695-96 n. 4, 101 S.Ct. 2587. In other words, a brief detention may be justified by circumstances unrelated to the apparent danger posed by the person detained; a search of that person requires additional justification. That ought to be the end of our case. The police detained Johnson (or tried to) but did not frisk him until his defiance of their instructions produced ample cause — the discovery of a weapon.
According to my colleagues, Summers is not dispositive because the police arrived at the house in Michigan with a search warrant, while the police arrived at the apartment in Illinois with only a suspicion that crime was afoot. That is a factual difference, to be sure, and one the Supreme Court mentioned, but does it justify a legal difference? A point of Summers, and other cases too, is that less is required to support a short detention than to support a search, and that one good justification for such a limited intrusion is risk: “Although no special danger to the police is suggested by the evidence in this record, the execution of a warrant to search for narcotics is the kind of transaction that may give rise to sudden violence or frantic efforts to conceal or destroy evidence.” 452 U.S. at 702, 101 S.Ct. 2587 (footnote omitted). Just so here, one would think. Whether to confine Summers to the execution of search warrants (as the majority does) or recognize the broader principle that risks during drug investigations justify precautionary stops depends on how well that latter approach fits with other parts of fourth amendment jurisprudence — and with the world at large.
Many cases in addition to Summers say that a particular search or seizure may be deemed reasonable because of risks that do not point a finger of suspicion at a particular person. Consider a few examples:
• After stopping a vehicle for a traffic infraction, the police may insist that the driver get out of the car, and that the passenger remain seated (or get out). The rationale for this seizure of both driver and passenger is the riskiness to officers of traffic stops. Pennsylvania v. Mimms, 434 U.S. 106, 98 S.Ct. 330, 54 L.Ed.2d 331 (1977); Maryland v. Wilson, 519 U.S. 408, 117 S.Ct. 882, 137 L.Ed.2d 41 (1997).
• After making a custodial arrest, the police may conduct a complete search of the person, including the contents of any packages he may be carrying, to ensure that none of the items is dangerous. United States v. Robinson, 414 U.S. 218, 94 S.Ct. 467, 38 L.Ed.2d 427 (1973). Cf. Knowles v. Iowa, — U.S. -, 119 S.Ct. 484, 142 L.Ed.2d 492 (1998).
*724• Police may set roadblocks and stop all cars in order to check drivers’ licenses, insurance, intoxication, and other factors that affect the risk posed to motorists. Michigan Department of State Police v. Sitz, 496 U.S. 444, 110 S.Ct. 2481, 110 L.Ed.2d 412 (1990). Cf. Delaware v. Prouse, 440 U.S. 648, 99 S.Ct. 1391, 59 L.Ed.2d 660 (1979).
• After entering a dwelling to make an arrest, police may conduct a visual “protective sweep” of other rooms to ensure that armed occupants there do not pose a risk. Maryland v. Buie, 494 U.S. 325, 110 S.Ct. 1093, 108 L.Ed.2d 276 (1990).
• School officials may conduct random urinalysis checks of junior high school athletes in light of the hazards drug use poses on the field. Vernonia School District 47J v. Acton, 515 U.S. 646, 115 S.Ct. 2386, 132 L.Ed.2d 564 (1995); see also Dimeo v. Griffin, 943 F.2d 679, 682-83 (7th Cir.1991) (en banc) (fourth amendment permits drug tests of jockeys and other horse race participants in light of the sport’s dangerousness).
The list could be extended (think of airport searches for weapons and explosives), and it has parallels in other parts of the law. For example, police may ask a suspect whether he has any weapons before administering Miranda warnings, in order to curtail the risk that firearms pose. New York v. Quarles, 467 U.S. 649, 104 S.Ct. 2626, 81 L.Ed.2d 550 (1984).
Doubtless each of these situations differs in some respect from our case. Mimms, Wilson, and Sitz involve cars; Robinson and Buie involve arrests; Acton involves youngsters; Dimeo involves horses; and so on. To my colleagues these differences show that the cases are irrelevant; to me these many differences show the scope of the principle that situational risks justify minimal law-enforcement demands, such as the stay-where-you-are-for-a-few-minutes order issued to Johnson. Knowles makes this very point when holding that persons in a car stopped for a traffic violation may not be subject to the same full search appropriate to a custodial arrest. As the Court observed in Knowles, the demands made in Mimms and Wilson, such as “get out of the car” or “remain where you are while we write a ticket” are modest compared with a full search, and accordingly justifiable if the officers face a modest danger. See also United States v. Sakyi, 160 F.3d 164 (4th Cir.1998) (circumstances making a traffic stop out of the ordinary justify a frisk as well as an order to get out of the car). Likewise the demand made of Johnson called for very little of his time, and the danger to the officers justifying the imposition was appreciable. See United States v. Place, 462 U.S. 696, 709-10 & n. 10, 103 S.Ct. 2637, 77 L.Ed.2d 110 (1983) (implying the propriety of stops that do not exceed 20 minutes, although eschewing a “rigid” time limit); ALI, Model Code of Pre-Arraignment Procedure § 110.2(1) (1975).
How hazardous must a police-citizen encounter be to justify an order to stand still (or get out of a car)? In Wilson the Supreme Court told us that it thought the risks of traffic stops sufficient because during 1994 officers suffered 5,762 assaults and 11 deaths during traffic pursuits or stops. 519 U.S. at 413, 117 S.Ct. 882, citing Federal Bureau of Investigation, Uniform Crime Reports: Law Enforcement Officers Killed and Assaulted 71, 33 (1994). The risk per stop was quite modest, given the large number of traffic stops in the whole United States over the course of a year — and the fact that some of the fatalities reported by the FBI surely came from high-speed traffic pursuits, rather than stops. (Data in Steven G. Brandi, In the Line of Duty: A Descriptive Analysis of Police Assaults and Accidents, 24 J.Crim. Justice 255 (1996), imply that two-thirds of the injuries in traffic cases are attributable to chases.) In 1997, the most recent year for which data are available, 7 officers died in traffic pursuits or stops, and 9 died while “investigating suspicious persons/eireum-stances”, the category that seems closest to the investigation of Apartment 7. Federal Bureau of Investigation, Uniform Crime Reports: Law Enforcement Officers Killed and Assaulted 29 (Table 15) (1998). During a 10-year period, 89 officers died as a result of traffic stops or pursuits, and 128 died during investigations. Another 250 died while making arrests (45 of these in the subcategory of *725drug-related arrests). There are far fewer drug arrests (or investigations) than traffic stops, so the risk to officers during drug investigations seems appreciably greater. Accord, Allen P. Bristow, Police Officer Shootings — A Tactical Evaluation, 54 J.Crim. L., Criminology & Police Sci. 93 (1963) (finding that suspects approached in buildings are substantially more dangerous to police than are suspects approached in cars). As for injuries: in 1997 traffic pursuits and stops produced 5,043, “investigating suspicious persons/circumstances” led to 5,446, and arrests (other than robbery or burglary) were associated with 8,373 injuries. Law Enforcement Officers Killed and Assaulted at 75 (Table 29). Detective Mathy, who led the team to Apartment 7, was wearing body armor; perhaps he had seen these tables.
Drugs and guns go together, and armed persons are tempted to use their weapons. Asking Johnson to stand still (or reenter the apartment) was a prudent precaution. Judge Evans writes that in a situation such as this police “will (and should) always do what is necessary to protect themselves.” 170 F.3d at 721. From this, and the fact that it was lawful to approach Apartment 7 in the first place, it follows that the demand made of Johnson was a “reasonable” seizure within the meaning of the fourth amendment. Judge Evans appeals to the principle that police who needlessly create danger can’t invoke exigent circumstances to justify what follows, but walking through the public spaces of an apartment house to see whether the occupants of an apartment are willing to talk can’t be put in that category — not without discarding the principle that the police don’t need “probable” (or any) cause to knock on a door and ask for cooperation. The police did not create the hazard; they just approached it and were entitled to protect themselves. What made this seizure reasonable was knowledge about the risks associated with the activity, not knowledge about Johnson personally. Officers must be entitled to act on systemic information. Time and again the Supreme Court holds that they may; that principle is no less applicable to drug investigations. Any cop on the beat who thinks that information about a person’s location and associates cannot justify a stop would be wrong. Dead wrong.