At about 5:00 P.M. on January 25, 1966, on Erie Street, S.E., three men assaulted and robbed James Warwick of a wallet, approximately $200 in cash, and two checks payable to Warwick’s employer, W. A. Dawson. Immediately after the attack on Warwick, a motorist, Donald R. Leigh, saw three Negro males running from the scene of the robbery. Though he did not then know that a crime had been committed, Leigh did notice that the three men jumped into a blue 1953 or 1954 Chevrolet hardtop. Subsequently, after he came to the aid of Warwick, the District of Columbia police arrived and Leigh gave them this description of the car and the men.

At about 5:20 P.M., the police broadcast a radio lookout for three male Negroes in a 1953 blue Chevrolet wanted in connection with the robbery and assault on Erie Street. Shortly thereafter Police Officer Bailey, who was stationed about 3.7 miles from Erie Street, spotted a blue 1954 Chevrolet hardtop, occupied by four Negro males, heading away from the direction of the crime. Unable to follow, the officer broadcast his observation, including the D.C. license plate number of the car. Police headquarters also issued a supplemental lookout describing the occupants of the car as Negro males in their late teens or early twenties wearing dark clothing.

Officer Stone and his partner, who were cruising in Officer Bailey’s area, heard both the initial lookout and the Bailey broadcast. Stone soon spotted the car Bailey had observed and followed it for several blocks while awaiting other squad cars summoned to the area. At around 5:40 P.M., at 14th and C Streets, N.E., about six miles from the robbery, Officer Stone’s squad car forced appellants’ car to the side of the road while two other police cars blocked it from the front and rear. Officer Stone approached appellants’ car, apparently with his gun in hand, and ordered the occupants to sit still and keep their hands in plain sight. He then opened the front door on the driver’s side and asked the driver, appellant Russell, for his driver’s permit. As the officer was returning the permit he noticed a wallet on the right side of the floor of the car. He then went to the other side of the car, opened the door, and retrieved the wallet. In it were the two checks payable to W. A. Dawson. The appellants were then frisked on the scene and taken to the precinct house where they were thoroughly searched. Each had between $55 and $91 in cash. They were then placed in a lineup, after which Warwick stated that the clothing of three of them, appellants Frye, Bailey and Oliver, resembled that worn by his three assailants.

*308Before trial there were three separate hearings on motions to suppress filed in behalf of the appellants. Three different judges heard the motions and decided that the evidence seized from the car and from appellants was admissible. At the ensuing trial before a fourth judge the Government offered in evidence the items seized, including the stolen wallet and the money found on each of the appellants. Appellants’ objection to the offer was overruled by the trial judge, appellants were convicted, and these appeals followed.

Essentially three issues are raised on appeal. First, all appellants contend that the seized stolen items should not have been admitted in evidence. Second, each urges that there should have been a directed verdict of acquittal for insufficient evidence in that each may have been the fourth occupant of the ear, seen by neither Warwick nor Leigh. Third, appellant Frye argues that he should have been permitted to testify at the hearing on his motion to suppress without waiving his Fifth Amendment right not to incriminate himself.

(1) Admissibility of the evidence.

The Fourth Amendment1 prohibits unreasonable searches and seizures, but it is well established that a search incident to a lawful arrest is permissible even without a warrant. Weeks v. United States, 232 U.S. 383, 392, 34 S.Ct. 341, 58 L.Ed. 652 (1914). This right unquestionably extends to a contemporaneous search of the car in the control of the accused at the time of arrest. Preston v. United States, 376 U.S. 364, 367, 84 S.Ct. 881, 11 L.Ed.2d 777 (1964). The arrest which justifies the search, however, is lawful only if the Fourth Amendment standard of probable cause is met. Beck v. State of Ohio, 379 U.S. 89, 91, 85 S.Ct. 223, 13 L.Ed.2d 142 (1964). And there must be probable cause before the search begins, for “a search is not to be made legal by what it turns up. In law it is good or bad when it starts and does not change character from its success.” United States v. DiRe, 332 U.S. 581, 595, 68 S.Ct. 222, 229, 92 L.Ed. 210 (1948). This is true even when the arrest or search is made pursuant to a warrant. Where, as here, the arrest is made without a warrant the standard is at least as high. Wong Sun v. United States, 371 U.S. 471, 479-480, 83 S.Ct. 407, 9 L.Ed.2d 441 (1963); Beck v. State of Ohio, supra, 379 U.S. at 96, 85 S.Ct. 223.

Though the arresting officers here did not think that a formal arrest took place until after the stolen wallet was spotted, seized and examined, the Government, as in Henry v. United States, 361 U.S. 98, 103, 80 S.Ct. 168, 4 L.Ed.2d 134 (1959), concedes that in law the arrest of the car’s passengers as well as its driver took place at the time Officer Stone approached the car, perhaps with gun drawn, and told appellants to sit still and keep their hands in plain sight. Even if the formal arrest was not made until after the search, the search will be upheld so long as there is probable cause for an arrest before the search is begun. See United States v. Gorman, 2 Cir., 355 F.2d 151, 159-160 (1965), cert. denied, 384 U.S. 1024, 86 S.Ct. 1962, 16 L.Ed.2d 1027 (1966). Accepting the Government’s concession, the question then is whether there was probable cause for an arrest when Officer Stone first approached appellants’ ear. We think that there was.

Probable cause is a plastic concept whose existence depends on the facts and circumstances of the particular case. It has been said that “ ‘[t]he substance of all the definitions’ of probable cause ‘is a reasonable ground for belief of guilt.’ ” Brinegar v. United States, 338 U.S. 160, 175, 69 S.Ct. 1302, 1310, 93 L.Ed. 1879 (1949). Much less evidence *309than is required to establish guilt is necessary. Draper v. United States, 358 U.S. 307, 311-312, 79 S.Ct. 329, 3 L.Ed. 2d 327 (1959). The standard is that of “a reasonable, cautious and prudent peace officer” and must be judged in the light of his experience and training. Bell v. United States, 102 U.S.App.D.C. 383, 387, 254 F.2d 82, 86, cert. denied, 358 U.S. 885, 79 S.Ct. 126, 3 L.Ed.2d 113 (1958). The police must have enough information to “warrant a man of reasonable caution in the belief” that a crime has been committed and that the person arrested has committed it. Carroll v. United States, 267 U.S. 132, 162, 45 S.Ct. 280, 288, 69 L.Ed. 543 (1925). See also Henry v. United States, supra, 361 U.S. at 102, 80 S.Ct. 168. A finding of probable cause depends on the “practical considerations of everyday life on which reasonable and prudent men, not legal technicians, act.” Brinegar v. United States, supra, 338 U.S. at 175, 69 S.Ct. 1310.

We think the police acted reasonably in this case. They knew that a robbery had been committed by three Negro men who had then escaped in a 12- or 13-year-old car of which they had an apparently accurate description. They also had a general description of its occupants. A car matching this description was then seen at a time and distance from the robbery consistent with its being the getaway car. It was effective law enforcement which enabled the police to cordon off the suspects’ car and avoid the hazards of a high speed chase. Clearly this was not an arrest for investigation of the kind we condemned in Gatlin v. United States, 117 U.S.App.D.C. 123, 326 F.2d 666 (1963). The police here were obviously quite certain they had the right men. And, most important, there is no suggestion of a roundup of innocent “suspects” following the robbery.

We do not think it significant that the vintage blue car stopped here had four occupants while the lookout was for three suspects. The police could reasonably suppose that a fourth man, serving as a lookout, had waited in the car while the other three perpetrated the robbery itself. Nor do we think it significant that a 1954 Chevrolet was stopped while the lookout was for a 1953 Chevrolet, for the record shows the two models are very similar. Even wider discrepancies between the description in the police lookout and the man arrested were present in Brown v. United States, 125 U.S.App.D.C. 43, 365 F.2d 976 (1966), where this court sustained the denial of a motion to suppress on facts similar to those of this case.

Had the police waited for more detailed information before making the arrests, the suspects’ identities might never have been ascertained or the car might have sped on into one of the neighboring jurisdictions. The “exigencies of the situation made [the police] course [of action] imperative.” See McDonald v. United States, 335 U.S. 451, 456, 69 S.Ct. 191, 193, 93 L.Ed. 153 (1948). As the Supreme Court said last term in Warden, Md. Penitentiary v. Hayden, 387 U.S. 294, 298-299, 87 S.Ct. 1642, 1646, 18 L.Ed.2d 782 (1967), “The Fourth Amendment does not require police officers to delay in the course of an investigation if to do so would gravely endanger their lives or the lives of others. Speed here was essential * * So too in our case was speed essential if the suspected perpetrators of the robbery and brutal assault were to be captured. The emergency character of these arrests weighs heavily in determining their reasonableness. “Common sense dictates, of course, that questions involving searches of motorcars or other things readily moved cannot be treated as identical to questions arising out of searches of fixed structures like houses. For this reason, what may be an unreasonable search of a house may be reasonable in the ease of a motorcar.” Preston v. United States, supra, 376 U.S. at 366-367, 84 S.Ct. at 883. See also Cooper v. State of California, 386 U.S. 58, 87 S.Ct. 788, 17 L.Ed.2d 730 (1967); Price v. United States, 121 U.S.App.D.C. 62, 348 F.2d 68, cert. denied, 382 U.S. 888, 86 S.Ct. 170, 15 L.Ed.2d 125 (1965). The element of flight in a vehicle from the scene of the *310crime tips the scales here in favor of probable cause. Carroll v. United States, supra; Brinegar v. United States, supra. Compare Johnson v. United States, 333 U.S. 10, 14-15, 68 S.Ct. 367, 92 L.Ed. 436 (1948); McDonald v. United States, supra, 335 U.S. at 454-456, 69 S.Ct. 191.

(2) Sufficiency of the evidence.

Each appellant relies on this court’s decision in Goodwin v. United States, 121 U.S.App.D.C. 9, 347 F.2d 793, cert. denied, 382 U.S. 920, 86 S.Ct. 298, 15 L.Ed.2d 234 (1965), to argue that the trial judge erred in not directing a verdict in his favor. The facts in Goodwin are much like those of this case. In Goodwin the four occupants of a getaway car were arrested about an hour after the robbery of a grocery store. A search of the car turned up the stolen money as well as the gun used to threaten the store’s proprietor. There, as here, the robbery itself was committed by only three men, but there the victim positively identified three of the four suspects as his assailants. The fourth, unidentified, defendant, Paul Vaughn, testified at trial and denied his participation in the crime. Nevertheless, all four were convicted. On appeal this court reversed Vaughn’s conviction on the ground that there was not enough evidence to establish “beyond a reasonable doubt that he had been the look-out man. Nor was [his] presence in the car at the time of the arrests sufficient, we think, to show that he was in possession of the recently stolen articles here involved.” 121 U.S.App.D.C. at 11, 347 F.2d at 795.

In the instant case there was not a positive identification made of any of the suspects, and each argues that he is the fourth man, standing in the shoes of Paul Vaughn. This argument is stressed particularly by appellant Russell. We think that the Goodwin case, though similar, is sufficiently different in two significant respects to require an affirmance of all four convictions in this case. First, the only suspect who was not even tentatively identified as a participant in the actual assault and robbery was the driver, Russell. That he was the driver, and not a passenger in the back seat as Vaughn was, makes his presence more consistent with his being a lookout than with his being an innocent passenger who joined the others after the crime was committed. Second, and more important, each of the appellants here was found with what was almost certainly part of the loot in his immediate possession. This was not true in Goodwin, where all the money was found in a plastic bag on the floor of the car and there was no evidence that Paul Vaughn was going to participate in its distribution. Though the cash here was fungible and could not be identified by serial number or denomination, unless the money found on each of the four appellants is cumulated, all the stolen funds cannot be accounted for. Thus the inference of guilt from the possession of recently stolen articles is much stronger here than it was in Goodwin, and is sufficiently strong, with the other circumstances pointing toward guilt, to require affirmance. Cf. Norman v. United States, 126 U.S.App.D.C. 387, 379 F.2d 164, cert. denied, 389 U.S. 886, 88 S.Ct. 167, 19 L.Ed.2d 186 (1967).

(3) Self-incrimination.

Appellant Frye’s attorney apparently intended to put his client on the stand at the hearing on his motion to suppress. He declined to do so, however, when the District Court judge ruled that whatever Frye said there could be used against him at trial. On appeal Frye argues that he should have been allowed to testify at the hearing without waiving his Fifth Amendment privilege not to incriminate himself. We agree.

It is the rule in this jurisdiction that the defendant can challenge the voluntariness of his confession outside the jury’s presence without waiving his privilege against self-incrimination. Wright v. United States, 102 U.S.App. D.C. 36, 45, 250 F.2d 4, 13 (1957). There is no sound reason why the rule should be different where the defendant challenges the admissibility of non-testimonial evidence at the hearing on his motion to suppress. The Government contends that Wright holds only that where the *311defendant successfully challenges the vol-untariness of his confession his testimony at the hearing is inadmissible at trial. We do not read this limitation into Wright. There we said, simply and unequivocally, that “[t]o substantiate a defendant’s contention that his confession was involuntary, it is generally necessary for him to take the stand. Since it is a matter of course that the defendant should be allowed to testify to the involuntariness of his confession without waiving his privilege against self-incrimination, 3 Wigmore, Evidence 345 (3d Ed. 1940), it follows that, when he requests it, he should be given a hearing without the jury.” Ibid.

The Government also relies on a series of cases2 from other circuits to argue that, whatever the Wright rule may be as to voluntariness hearings, the law concerning the use of the defendant’s testimony taken at a Rule 41(e), Fed.R.Crim.P., hearing to suppress evidence is clear — if the defendant testifies and loses on his motion to suppress, his testimony can be used substantively against him at trial. The Government concedes that under this rule “in some cases a defendant wishing to challenge the legality of a search may find himself, on the horns of a dilemma, having to incriminate himself to assert his claim of illegality,” but argues that this is not a dilemma of the Government’s making.

Most of the cases cited by the Government in support of this proposition antedate the decision in Jones v. United States, 362 U.S. 257, 80 S.Ct. 725, 4 L.Ed.2d 697 (1960), where the Supreme Court rejected a whole line of Court of Appeals cases and held that a defendant had standing under Rule 41(e) to challenge the legality of a seizure of narcotics without admitting ownership. Prior to Jones the defendant’s dilemma was not of the Government’s making either, but the Court nonetheless dissolved it, holding that the exercise of a constitutional right cannot be embarrassed by placing the defendant in such a position.

We think the implications and logic of Jones require that the testimony of the defendant taken at the suppression hearing, although admissible for impeachment, cannot be used affirmatively against him at trial. 362 U.S. at 261-265, 80 S.Ct. 725. The Fourth Amendment protection against unreasonable searches and seizures is not a stepchild of the Constitution,3 nor is the privilege against self-incrimination. There is no reason why the defendant should be forced to risk waiving the one in order to raise the other. Compare Garrity v. State of New Jersey, 385 U.S. 493, 87 S.Ct. 616, 17 L.Ed.2d 562 (1967). Were we to adopt the Government’s rule many unlawful searches and arrests would go unchallenged.

We do not, however, reverse in this case because, applying the rule of Chapman v. State of California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), we find beyond a reasonable doubt that appellant Frye was not prejudiced. The only issue at the hearing on his motion to suppress was whether the police had probable cause to make an arrest. In his brief on appeal Frye argues that he would have testified on the question of just when the arrest occurred. Assuming that the suspect’s frame of mind is relevant to a determination of this issue, *312Frye could not have been prejudiced since we have found that there was probable cause at the earliest possible time the arrest can in law be said to have been made. At oral argument, Frye’s attorney stated that, at the hearing on the motion to suppress, Frye would have testified that he joined the others after the robbery occurred and played no part in it. But clearly such evidence is not relevant to the question whether the police had probable cause to arrest the occupants of the car. If Frye wished to testify as to his non-participation, he should have done so at trial.

We have considered the other points urged on appeal and find them to be without merit.

Affirmed.

(concurring).

I concur in parts two and three of the court’s opinion. I do not join in part one because I cannot accept the majority’s evaluation of the police conduct. While I agree that the officers acted reasonably, we should not refashion what was done in order to fit it to the frame of traditional analysis of arrests. The court skirts the problems these cases pose by accepting the prosecutor’s concession that an arrest took place when the car was first stopped, only to justify that “arrest” by stretching probable cause. Probable cause may be a plastic concept, but not for pressing into any and all molds.

The critical facts, as I see them, are these. Three Negro men committed a brutal robbery at 5:00 p. m., and were seen jumping into what was described as a 1953 blue Chevrolet. A police radio lookout at 5:20 alerted police officers to watch for them. Thereafter, a traffic policeman, some 3.7 miles from the scene of the crime, saw a blue 1954 Chevrolet occupied by four Negro males coming from that general direction. He called in his observation, which was broadcast. The car was sighted and followed by Officer Stone, who had heard both the original lookout and a supplement which described the suspects as young and wearing dark clothing and advised that a brown wallet had been stolen. Officer Stone and his partner trailed appellants’ car for some distance, apparently waiting for their superior, Detective Richardson, to get to the scene. Finally, as other police cars arrived, Officer Stone pulled appellants’ car to the curb, using his siren for about a block. On approaching the car, he told appellants to sit still and keep their hands in sight. In his testimony, he said he could not remember whether he had drawn his gun. He asked for and examined appellant Russell’s driver’s license. On returning it, he saw a brown wallet on the floor protruding from under the front seat. He asked whose it was, and when no one answered he went around to the other side of the car and seized it, giving it to Detective Richardson who searched it and found it to be the robbery victim's. Appellants were then arrested.

The court finds probable cause to arrest as of the time the car was stopped. Yet at that time all the police knew was that the car and its occupants were a very rough match to those involved in a robbery nearly four miles away from where the car was first sighted. It was rush hour,1 with traffic at its heaviest in a poor area of the city, where old cars are relatively numerous. There was nothing suspicious about the way the occupants of the car behaved. I am unwilling to say that any group of three or four Negroes who happened to be riding in an old Chevrolet on that day in a fifty square mile area can without more be searched, booked, and stigmatized by an arrest record regardless of whether they could explain the whole thing away if given an immediate chance. Yet all of these are permissible police actions, if there was probable cause for arrest.2

*313When we deal with arrests without warrant there is some scope for saying that events or conditions justified the police officers in exercising their discretion. As the majority aptly points out, probable cause depends on the “practical considerations of everyday life on which reasonable and prudent men, not legal technicians, act.” Brinegar v. United States, 338 U.S. 160, 175, 69 S.Ct. 1302, 1310 (1949). The standard is indeed that of “a reasonable, cautious and prudent peace officer” who must be judged in the light of his experience and training. Bell v. United States, 102 U.S.App.D.C. 383, 387, 254 F.2d 82, 86, cert. denied, 358 U.S. 885, 79 S.Ct. 126, 3 L.Ed.2d 113 (1958). But here the record is plain that the police did not intend to arrest these appellants when they were stopped.

Two officers testified at trial. Officer Stone, who first approached the car, was asked, “At this time if the people in the car had wanted to leave, would they have been able to leave?” He answered, “As far as I am concerned, they could. I had not placed them under arrest.” Detective Richardson, who made the formal arrest after examining the stolen wallet, was asked, “when you arrived on the scene were these men under arrest?” His reply was, “To my knowledge, no.” Of course, police testimony is not dispositive on the issue of when an arrest is made.3 While the matter was not probed at trial, it is fair to assume and for present purposes I am willing to assume that the officers intended to detain the car while they asked questions like “where have you been” and examined Russell’s driver’s license. What is clear from their testimony that appellants had not been placed under arrest, and hence were free to go, is that they did not feel they had sufficient basis for taking the occupants to the station house for booking at the time they pulled the appellants’ car to the curb. The doubts of the reasonable and prudent police officers who were on this scene augment and confirm my own.

The court says that at the time of the stop the police were “quite certain they had the right men.” This is not supported by the record. If they were so sure they had the right men, why would both officers testify that at the time of the stop the men were free to go? This is not a question of informal rather than formal arrest. The police simply did not know enough to justify arrest. Other policemen stopped and impounded at least one other old Chevrolet, because that car was first shown to Mr. Leigh, who witnessed the getaway. Athough not gone into at trial, there are hints that still other ears were stopped. To say that these officers stopping this car knew they had their men is to implant a certitude when to me the only thing that is certain is that the officers were by no means certain. In fact the only element of perhaps unusual policework, blocking off the car by police cars, apparently occurred by accident rather than design. Another car coming to the scene happened to round the corner just as Officer Stone was making the stop.

The real difficulty in this case, in my view, is that rigorous focus on the facts that I think are clear in the record requires us to do more than merely dip a toe in the water in setting out standards for detentions where there are no grounds for arrest.4 Investigation is too important and recurring a police function to have good policework commended only after judicial brushwork touches up the way situations are handled in the real world. Investigation is subject to its own types of abuses, but these should *314be guarded against as such and not dealt with by manipulating the already overused concepts of arrest and probable cause.

I do not think an arrest took place as a matter of law when the car was stopped. Nor do I think the legal situation affected by a concession proffered by the Government for tactical reasons.5 What the police did was stop a car to check-out what everyone concedes were suspicious circumstances. The occupants were no more arrested at that point than are the drivers who are stopped by a roadblock thrown up to prevent a kidnapper from fleeing with a child.6 A mere stop— though it also must have some justification in the facts — is not necessarily an “arrest” that triggers the requirement that the person be taken to a magistrate “as quickly as possible” as Rule 5 commands when an arrest is made. Certainly, if nothing had come of the matter, neither the police nor the persons stopped would ever have supposed there had been an arrest, and the law does not say otherwise. What the police did was to act reasonably to bring a situation under control ; they had no way of knowing whether the car would leave the jurisdiction, and once the occupants scattered it would be nearly impossible to reassemble them again. As Judge McGowan pointed out in Dorsey v. United States, 125 U.S.App.D.C. 355, 358, 372 F.2d 928, 931 (1967): “If policemen are to serve any purpose of detecting and preventing crime by being out on the streets at all, they must be able to take a closer look at challenging situations as they encounter them.”

Appointed counsel for the appellants have divided for tactical reasons on whether the stop constituted an arrest. Those who say it did lean on Henry v. United States, 361 U.S. 98, 80 S.Ct. 168, 4 L.Ed.2d 134 (1959) where the Supreme Court found an arrest “when the officers interrupted the two men and restricted their liberty of movement,” 361 U.S. at 103, 80 S.Ct. at 171. Henry is extended much too far when argued as a holding that every stop is an arrest. Henry turned on its particular facts,7 and the following year in Rios v. United States, 364 U.S. 253, 262, 80 S.Ct. 1431, 4 L.Ed. 2d 1688 (1960) the Court indicated that no arrest is involved when the detention of the person or vehicle involved is only what is incident to a brief on-the-street questioning for the purpose of clearing up suspicious circumstances. The same approach may underlie the distinction drawn in Miranda between custodial interrogation and general on-the-scene investigation.

I realize that this case was not a usual brief on-the-street detention, and that it is a hard case on its facts. Two items stand out — Officer Stone’s command that the occupants of the car sit still and keep their hands in sight, and the number of police cars on the scene, at least four. Whether there has been an arrest turns on whether there has been an imposition of custody, and this is a determination made after examining both the objective circumstances and the subjective feeling those circumstances are likely to evoke. Seals v. United States, 117 U.S.App.D.C. 79, 325 F.2d 1006 (1963), cert. denied, 376 U.S. 964, 84 S.Ct. 1123, 11 L.Ed.2d 982 (1964); Kelley v. United States, 111 U.S.App.D.C. 396, 298 F.2d 310 (1961); Coleman v. United States, 111 U.S.App. D.C. 210, 295 F.2d 555 (1961) (en banc), cert. denied, 369 U.S. 813, 82 S.Ct. 689, 7 L.Ed.2d 613 (1962).

I am willing to assume that in this case, regardless of what the officers intended, the persons who were stopped supposed — and reasonably so — that they *315were not free to go until they responded to the officer’s queries (even if only to refuse to answer). There is reason for overriding concern with a person’s probable state of mind in a case where the issue is the admissibility of a statement,— because such a statement cannot be said to be truly voluntary unless his state of mind embraces an awareness that he is free to speak or not, Townsend v. Sain, 372 U.S. 293, 307-309, 83 S.Ct. 745, 9 L.Ed.2d 770 (1963). Subjective elements have a lesser role in evaluating whether a stop constitutes an arrest. We are not dealing here with psychological gamesmanship staged in the backroom of the police station. As a society, we routinely expect police officers to risk their lives in apprehending dangerous people. We should not bicker if in bringing potentially dangerous situations under control they issue commands and take precautions which reasonable men are warranted in taking. That is particularly true when, as here, rush hour traffic conditions mean that any fight or flight will endanger the lives of bystanders who have not signed on to risk getting shot.

It is an unpleasant and doubtless anxiety-causing experience to feel that the police are telling you to stay put. As citizens in a free society, we are generally free to do as we please — within broad limits — and any limitations on our freedom must be closely scrutinized. Still, in the stop situation, I think it unwise to focus on whether the detained person feels “free to go.” As a realistic matter, there can be no casual disregard of on-the-street police questioning. People generally do not know and usually do not care whether they have a lawful right to walk away without regard for the pres-enee of the policeman. Usually they are perfectly willing to cooperate with the authorities. In other cases they would prefer to clear a matter up right away rather than seem suspicious. People who are stopped for some questions generally stay stopped until the police indicate they can go. If states of mind at the moment could be discerned, most people would say, I think, that while they were certainly not under arrest, neither were they free to go.

It cannot be a decisive inquiry whether the person involved was “free to go.” It requires too much of reasonable men that they try to choose between letting a car containing four possibly desperate and dangerous men leave the jurisdiction unquestioned or else stop it in the middle of rush hour without thought of the safety of themselves or others. If stopping them safely necessitates an arrest as a matter of law, we foster an anomaly because the police need more than suspicion to justify an arrest. It would erect barriers to investigation of precisely those crimes about which society-at-large is most concerned. On the other end of the spectrum, however, if detentions for interrogation are broadly upheld on the theory that the suspect could always have exercised a theoretical right to go, this may be too little protection against capricious interruption and investigative abuses.

In this context, it seems to me that the key question is whether the constraints applied were protective of the police and bystanders, or were custodial.8 In the overwhelming number of cases, the police do nothing more than walk or drive up and ask questions. That is no arrest.9 *316On the other hand, if the police approach a well-known pacifist with drawn guns, statements that they were in danger are simply not believable and the validity of the detention must be governed by standards controlling arrests. Obviously, the hard questions are in the middle, and can only be evaluated by looking, as in so many other Fourth Amendment questions, to the reasonableness of the police conduct. Commanding four robbery suspects to keep their hands in sight is eminently reasonable and does not, I think, automatically constitute custodial detention.

I do not think the cases of these appellants force us to face the hard question whether the law has crystallized to the point that a detention in the face of a specific refusal to answer questions and a request to go is necessarily an arrest, constitutionally permissible only when there is true probable cause.10

I join the majority in affirming these convictions because I think, as did the officers on the scene, that seeing the wallet stuffed under the floorboard of the front seat confirmed suspicion into probable cause sufficient to make the search reasonable. The grounds for suspicion were reasonable, and they justified the interruption of travel, but I do not think they, without more, conveyed a right to search the car. The police knew, however, that a brown leather wallet had been stolen in this robbery, and Officer Stone saw such a wallet in a highly unusual place while he was examining Russell’s driver’s license. He asked whose it was, and no one answered. This extra item it seems to me established probable cause for an arrest and justification for search of the automobile and the occupants.

The only other question of substance is whether the search and seizure are invalid because the arrest did not take place until afterwards, even though probable cause justifying it preceded. It seems to me that the basis of the search is made out by the existence of probable cause for arrest whether or not an arrest is made prior to the search.11