with whom Justice Brennan and Justice Marshall join as to Parts I and II, dissenting.
A sobriety checkpoint is usually operated at night at an unannounced location. Surprise is crucial to its method. The test operation conducted by the Michigan State Police and the Saginaw County Sheriff’s Department began shortly after midnight and lasted until about 1 a.m. During that period, the 19 officers participating in the operation made two arrests and stopped and questioned 124 other unsuspecting and innocent drivers.1 It is, of course, not known how many arrests would have been made during that period if those officers had been engaged in normal patrol activities. However, the findings of the trial court, based on an extensive record and affirmed by the Michigan Court of Appeals, indicate that the net effect of sobriety checkpoints on traffic safety is infinitesimal and possibly negative.
Indeed, the record in this case makes clear that a decision holding these suspicionless seizures unconstitutional would not impede the law enforcement community’s remarkable progress in reducing the death toll on our highways.2 Be*461cause the Michigan program was patterned after an older program in Maryland, the trial judge gave special attention to that State’s experience. Over a period of several years, Maryland operated 125 checkpoints; of the 41,000 motorists passing through those checkpoints, only 143 persons (0.3%) were arrested.3 The number of man-hours devoted to these *462operations is not in the record, but it seems inconceivable that a higher arrest rate could not have been achieved by more conventional means.4 Yet, even if the 143 checkpoint arrests were assumed to involve a net increase in the number of drunken driving arrests per year, the figure would still be insignificant by comparison to the 71,000 such arrests made by Michigan State Police without checkpoints in 1984 alone. See App. to Pet. for Cert. 97a.
Any relationship between sobriety checkpoints and an actual reduction in highway fatalities is even less substantial than the minimal impact on arrest rates. As the Michigan Court of Appeals pointed out: “Maryland had conducted a study comparing traffic statistics between a county using checkpoints and a control county. The results of the study showed that alcohol-related accidents in the checkpoint county decreased by ten percent, whereas the control county saw an eleven percent decrease; and while fatal accidents in the control county fell from sixteen to three, fatal accidents in the checkpoint county actually doubled from the prior year.” 170 Mich. App. 433, 443, 429 N. W. 2d 180, 184 (1988).
In light of these considerations, it seems evident that the Court today misapplies the balancing test announced in Brown v. Texas, 443 U. S. 47, 50-51 (1979). The Court overvalues the law enforcement interest in using sobriety checkpoints, undervalues the citizen’s interest in freedom from random, unannounced investigatory seizures, and mistakenly assumes that there is “virtually no difference” between a routine stop at a permanent, fixed checkpoint and a *463surprise stop at a sobriety checkpoint. I believe this case is controlled by our several precedents condemning suspicionless random stops of motorists for investigatory purposes. Delaware v. Prouse, 440 U. S. 648 (1979); United States v. Brignoni-Ponce, 422 U. S. 873 (1975); United States v. Ortiz, 422 U. S. 891 (1975); Almeida-Sanchez v. United States, 413 U. S. 266 (1973); cf. Carroll v. United States, 267 U. S. 132, 153-154 (1925).
I
There is a critical difference between a seizure that is preceded by fair notice and one that is effected by surprise. See Wyman v. James, 400 U. S. 309, 320-321 (1971); United States v. Martinez-Fuerte, 428 U. S. 543, 559 (1976); Michigan v. Tyler, 436 U. S. 499, 513-514 (1978) (Stevens, J., concurring in part and concurring in judgment). That is one reason why a border search, or indeed any search at a permanent and fixed checkpoint, is much less intrusive than a random stop. A motorist with advance notice of the location of a permanent checkpoint has an opportunity to avoid the search entirely, or at least to prepare for, and limit, the intrusion on her privacy.
No such opportunity is available in the case of a random stop or a temporary checkpoint, which both depend for their effectiveness on the element of surprise. A driver who discovers an unexpected checkpoint on a familiar local road will be startled and distressed. She may infer, correctly, that the checkpoint is not simply “business as usual,” and may likewise infer, again correctly, that the police have made a discretionary decision to focus their law enforcement efforts upon her and others who pass the chosen point.
This element of surprise is the most obvious distinction between the sobriety checkpoints permitted by today’s majority and the interior border checkpoints approved by this Court in Martinez-Fuerte. The distinction casts immediate doubt upon the majority’s argument, for Martinez-Fuerte is the only case in which we have upheld suspicionless seizures *464of motorists. But the difference between notice and surprise is only one of the important reasons for distinguishing between permanent and mobile checkpoints. With respect to the former, there is no room for discretion in either the timing or the location of the stop—it is a permanent part of the landscape. In the latter case, however, although the checkpoint is most frequently employed during the hours of darkness on weekends (because that is when drivers with alcohol in their blood are most apt to be found on the road), the police have extremely broad discretion in determining the exact timing and placement of the roadblock.5
There is also a significant difference between the kind of discretion that the officer exercises after the stop is made. A check for a driver’s license, or for identification papers at an immigration checkpoint, is far more easily standardized than is a search for evidence of intoxication. A Michigan officer who questions a motorist at a sobriety checkpoint has virtually unlimited discretion to detain the driver on the basis *465of the slightest suspicion. A ruddy complexion, an unbuttoned shirt, bloodshot eyes, or a speech impediment may suffice to prolong the detention. Any driver who had just consumed a glass of beer, or even a sip of wine, would almost certainly have the burden of demonstrating to the officer that his or her driving ability was not impaired.6
Finally, it is significant that many of the stops at permanent checkpoints occur during daylight hours, whereas the sobriety checkpoints are almost invariably operated at night. A seizure followed by interrogation and even a cursory search at night is surely more offensive than a daytime stop that is almost as routine as going through a tollgate. Thus we thought it important to point out that the random stops at issue in Ortiz frequently occurred at night. 422 U. S., at 894.
These fears are not, as the Court would have it, solely the lot of the guilty. See ante, at 452. To be law abiding is not necessarily to be spotless, and even the most virtuous can be unlucky. Unwanted attention from the local police need not be less discomforting simply because one’s secrets are not the stuff of criminal prosecutions. Moreover, those who have found—by reason of prejudice or misfortune—that encounters with the police may become adversarial or unpleasant without good cause will have grounds for worrying at any stop designed to elicit signs of suspicious behavior. Being stopped by the police is distressing even when it should not be terrifying, and what begins mildly may by happenstance turn severe.
For all these reasons, I do not believe that this case is analogous to Martinez-Fuerte. In my opinion, the sobriety checkpoints are instead similar to—and in some respects more intrusive than—the random investigative stops that the Court held unconstitutional in Brignoni-Ponce and Prouse. In the latter case the Court explained:
*466“We cannot agree that stopping or detaining a vehicle on an ordinary city street is less intrusive than a roving-patrol stop on a major highway and that it bears greater resemblance to a permissible stop and secondary detention at a checkpoint near the border. In this regard, we note that Brignoni-Ponce was not limited to roving-patrol stops on limited-access roads, but applied to any roving-patrol stop by Border Patrol agents on any type of roadway on less than reasonable suspicion. See 422 U. S., at 882-883; United States v. Ortiz, 422 U. S. 891, 894 (1975). We cannot assume that the physical and psychological intrusion visited upon the occupants of a vehicle by a random stop to check documents is of any less moment than that occasioned by a stop by border agents on roving patrol. Both of these stops generally entail law enforcement officers signaling a moving automobile to pull over to the side of the roadway, by means of a possibly unsettling show of authority. Both interfere with freedom of movement, are inconvenient, and consume time. Both may create substantial anxiety.” 440 U. S., at 657.
We accordingly held that the State must produce evidence comparing the challenged seizure to other means of law enforcement, so as to show that the seizure
“is a sufficiently productive mechanism to justify the intrusion upon Fourth Amendment interests which such stops entail. On the record before us, that question must be answered in the negative. Given the alternative mechanisms available, both those in use and those that might be adopted, we are unconvinced that the incremental contribution to highway safety of the random spot check justifies the practice under the Fourth Amendment.” Id., at 659.
*467II
The Court, unable to draw any persuasive analogy to Martinez-Fuerte, rests its decision today on application of a more general balancing test taken from Brown v. Texas, 443 U. S. 47 (1979). In that case the appellant, a pedestrian, had been stopped for questioning in an area of El Paso, Texas, that had “a high incidence of drug traffic” because he “looked suspicious. ” Id., at 49. He was then arrested and convicted for refusing to identify himself to police officers. We set aside his conviction because the officers stopped him when they lacked any reasonable suspicion that he was engaged in criminal activity. In our opinion, we stated:
“Consideration of the constitutionality of such seizures involves a weighing of the gravity of the public concerns served by the seizure, the degree to which the seizure advances the public interest, and the severity of the interference with individual liberty.” Id., at 50-51.
The gravity of the public concern with highway safety that is implicated by this case is, of course, undisputed.7 *468Yet, that same grave concern was implicated in Delaware v. Prouse. Moreover, I do not understand the Court to have placed any lesser value on the importance of the drug problem implicated in Brown v. Texas or on the need to control the illegal border crossings that were at stake in Almeida-Sanchez and its progeny.8 A different result in this case must be justified by the other two factors in the Brown formulation.
As I have already explained, I believe the Court is quite wrong in blithely asserting that a sobriety checkpoint is no more intrusive than a permanent checkpoint. In my opinion, unannounced investigatory seizures are, particularly when *469they take place at night, the hallmark of regimes far different from ours;9 the surprise intrusion upon individual liberty is not minimal. On that issue, my difference with the Court may amount to nothing less than a difference in our respective evaluations of the importance of individual liberty, a serious, albeit inevitable, source of constitutional disagreement.10 On the degree to which the sobriety checkpoint seizures advance the public interest, however, the Court’s position is wholly indefensible.
The Court’s analysis of this issue resembles a business decision that measures profits by counting gross receipts and ignoring expenses. The evidence in this case indicates that sobriety checkpoints result in the arrest of a fraction of one percent of the drivers who are stopped,11 but there is absolutely no evidence that this figure represents an increase over the number of arrests that would have been made by using the same law enforcement resources in conventional patrols.12 Thus, although the gross number of arrests is more *470than zero, there is a complete failure of proof on the question whether the wholesale seizures have produced any net advance in the public interest in arresting intoxicated drivers.
Indeed, the position adopted today by the Court is not one endorsed by any of the law enforcement authorities to whom the Court purports to defer, see ante, at 453-454. The Michigan police do not rely, as the Court does, ante, at 454-455, on the arrest rate at sobriety checkpoints to justify the stops made there. Colonel Hough, the commander of the Michigan State Police and a leading proponent of the checkpoints, admitted at trial that the arrest rate at the checkpoints was “very low.” 1 Record 87. Instead, Colonel Hough and the State have maintained that the mere threat of such arrests is sufficient to deter drunken driving and so to reduce the accident rate.13 The Maryland police officer who testified *471at trial took the same position with respect to his State’s program.14 There is, obviously, nothing wrong with a law enforcement technique that reduces crime by pure deterrence without punishing anybody; on the contrary, such an approach is highly commendable. One cannot, however, prove its efficacy by counting the arrests that were made. One must instead measure the number of crimes that were avoided. Perhaps because the record is wanting, the Court simply ignores this point.
The Court’s sparse analysis of this issue differs markedly from Justice Powell’s opinion for the Court in Martinez-Fuerte. He did not merely count the 17,000 arrests made at the San Clemente checkpoint in 1973, 428 U. S., at 554; he also carefully explained why those arrests represented a net benefit to the law enforcement interest at stake.15 Common *472sense, moreover, suggests that immigration checkpoints are more necessary than sobriety checkpoints: There is no reason why smuggling illegal aliens should impair a motorist’s driving ability, but if intoxication did not noticeably affect driving ability it would not be unlawful. Drunken driving, unlike smuggling, may thus be detected absent any checkpoints. A program that produces thousands of otherwise impossible arrests is not a relevant precedent for a program that produces only a handful of arrests which would be more easily obtained without resort to suspicionless seizures of hundreds of innocent citizens.16
*473III
The most disturbing aspect of the Court’s decision today is that it appears to give no weight to the citizen’s interest in freedom from suspicionless unannounced investigatory seizures. Although the author of the opinion does not reiterate his description of that interest as “diaphanous,” see Delaware v. Prouse, 440 U. S., at 666 (Rehnquist, J., dissenting), the Court’s opinion implicitly adopts that characterization. On the other hand, the Court places a heavy thumb on the law enforcement interest by looking only at gross receipts instead of net benefits. Perhaps this tampering with the scales of justice can be explained by the Court’s obvious concern about the slaughter on our highways and a resultant tolerance for policies designed to alleviate the problem by “setting an example” of a few motorists. This possibility prompts two observations.
First, my objections to random seizures or temporary checkpoints do not apply to a host of other investigatory procedures that do not depend upon surprise and are unquestionably permissible. These procedures have been used to address other threats to human life no less pressing than the threat posed by drunken drivers. It is, for example, common practice to require every prospective airline passenger, or every visitor to a public building, to pass through a metal detector that will reveal the presence of a firearm or an explosive. Permanent, nondiscretionary checkpoints could be used to control serious dangers at other publicly operated facilities. Because concealed weapons obviously represent one such substantial threat to public safety,17 I would suppose *474that all subway passengers could be required to pass through metal detectors, so long as the detectors were permanent and every passenger was subjected to the same search.18 Likewise, I would suppose that a State could condition access to its toll roads upon not only paying the toll but also taking a uniformly administered breathalyzer test. That requirement might well keep all drunken drivers off the highways that serve the fastest and most dangerous traffic. This procedure would not be subject to the constitutional objections that control this case: The checkpoints would be permanently fixed, the stopping procedure would apply to all users of the toll road in precisely the same way, and police officers would not be free to make arbitrary choices about which neighborhoods should be targeted or about which individuals should be more thoroughly searched. Random, suspicionless seizures designed to search for evidence of firearms, drugs, or intoxication belong, however, in a fundamentally different category. These seizures play upon the detained individual’s reasonable expectations of privacy, injecting a suspicionless search into a context where none would normally occur. The imposition that seems diaphanous today may be intolerable tomorrow. See Boyd v. United States, 116 U. S. 616, 635 (1886).
*475Second, sobriety checkpoints are elaborate, and disquieting, publicity stunts. The possibility that anybody, no matter how innocent, may be stopped for police inspection is nothing if not attention getting. The shock value of the checkpoint program may be its most effective feature: Lieutenant Cotten of the Maryland State Police, a defense witness, testified that “the media coverage . . . has been absolutely overwhelming .... Quite frankly we got benefits just from the controversy of the sobriety checkpoints.”19 In*476sofar as the State seeks to justify its use of sobriety checkpoints on the basis that they dramatize the public interest in the prevention of alcohol-related accidents, the Court should heed Justice Scalia’s comment upon a similar justification for a drug screening program:
“The only plausible explanation, in my view, is what the Commissioner himself offered in the concluding sentence of his memorandum to Customs Service employees announcing the program: ‘Implementation of the drug screening program would set an important example in our country’s struggle with this most serious threat to our national health and security.’ App. 12. Or as respondent’s brief to this Court asserted: ‘if a law enforcement agency and its employees do not take the law seriously, neither will the public on which the agency’s effectiveness depends.’ Brief for Respondent 36. What better way to show that the Government is serious about its ‘war on drugs’ than to subject its employees on the front line of that war to this invasion of their privacy and affront to their dignity? To be sure, there is only a slight chance that it will prevent some serious public harm resulting from Service employee drug use, but it will show to the world that the Service is ‘clean,’ and— most important of all—will demonstrate the determination of the Government to eliminate this scourge of our society! I think it obvious that this justification is unacceptable; that the impairment of individual liberties cannot be the means of making a point; that symbolism, *477even symbolism for so worthy a cause as the abolition of unlawful drugs, cannot validate an otherwise unreasonable search.” Treasury Employees v. Von Raab, 489 U. S. 656, 686-687 (1989) (dissenting opinion).
This is a case that is driven by nothing more than symbolic state action—an insufficient justification for an otherwise unreasonable program of random seizures. Unfortunately, the Court is transfixed by the wrong symbol—the illusory prospect of punishing countless intoxicated motorists—when it should keep its eyes on the road plainly marked by the Constitution.
I respectfully dissent.
The 19 officers present at the sole Michigan checkpoint were not the standard detail; a few were observers. Nevertheless, the standard plan calls for having at least 8 and as many as 12 officers on hand. 1 Record 82-83.
The fatality rate per 100 million miles traveled has steadily declined from 5.2 in 1968 to 2.3 in 1988. During the same span, the absolute number of fatalities also decreased, albeit less steadily, from more than 52,000 in 1968 to appoximately 47,000 in 1988. U. S. Dept. of Transportation, National Highway Traffic Safety Administration, Fatal Accident Report*461ing System 1988, Ch. 1, p. 6 (Dec. 1989) (hereinafter Fatal Accident Reporting System 1988).
Alcohol remains a substantial cause of these accidents, but progress has been made on this front as well:
“Since 1982, alcohol use by drivers in fatal crashes has steadily decreased. The proportion of all drivers who were estimated to have been legally intoxicated ([blood alcohol concentration] of .10 or greater) dropped from 30% in 1982 to 24.6% in 1988. The reduction from 1982-1988 is 18%.
“The proportion of fatally injured drivers who were legally intoxicated dropped from 43.8% in 1982 to 37.5% in 1988—a 14% decrease.
“During the past seven years, the proportion of drivers involved in fatal crashes who were intoxicated decreased in all age groups. The most significant drop continues to be in the 15 to 19 year old age group. In 1982, NHTSA estimated that 28.4% of these teenaged drivers in fatal crashes were drunk, compared with 18.3% in 1988.” Id., Overview, p. 2.
All of these improvements have been achieved despite resistance—now ebbing at last—to the use of airbags and other passive restraints, improvements that would almost certainly result in even more dramatic reductions in the fatality rate. Indeed, the National Highway Traffic Safety Administration estimates that an additional 5,000 lives per year would be saved if the 21 States without mandatory safety belt usage laws were to enact such legislation—even though only 50% of motorists obey such laws. Id., Overview, p. 4, Ch. 2, p. 13.
App. to Pet. for Cert. 80a-81a. The figures for other States are roughly comparable. See, e. g., State ex rel. Ekstrom v. Justice Ct., 136 Ariz. 1, 2, 663 P. 2d 992, 993 (1983) (5,763 cars stopped, 14 persons arrested for drunken driving); Ingersoll v. Palmer, 43 Cal. 3d 1321, 1327, 743 P. 2d 1299, 1303 (1987) (233 vehicles screened, no arrests for drunken driving); State v. Garcia, 481 N. E. 2d 148, 150 (Ind. App. 1985) (100 ears stopped, seven arrests for drunken driving made in two hours of operation); State v. McLaughlin, 471 N. E. 2d 1125, 1137 (Ind. App. 1984) (115 cars stopped, three arrests for drunken driving); State v. Deskins, 234 Kan. 529, 545, 673 P. 2d 1174, 1187 (1983) (Prager, J., dissenting) (2,000 to 3,000 vehicles stopped, 15 arrests made, 140 police man-hours consumed); Commonwealth v. Trumble, 396 Mass. 81, 85, 483 N. E. 2d 1102, 1105 *462(1985) (503 cars stopped, eight arrests, 13 participating officers); State v. Koppel, 127 N. H. 286, 288, 499 A. 2d 977, 979 (1985) (1,680 vehicles stopped, 18 arrests for driving while intoxicated).
“The then sheriffs of Macomb County, Kalamazoo County, and Wayne County all testified as to other means used in their counties to combat drunk driving and as to their respective opinions that other methods currently in use, e. g., patrol cars, were more effective means of combating drunk driving and utilizing law enforcement resources than sobriety checkpoints.” 170 Mich. App. 433, 443, 429 N. W. 2d 180, 184 (1988).
The Michigan plan provides that locations should be selected after consideration of “previous alcohol and drug experience per time of day and day of week as identified by arrests and/or Michigan Accident Location Index data,” App. to Pet. for Cert. 148a, and that “specific site selection” should be based on the following criteria:
“1. Safety of the location for citizens and law enforcement personnel. The site selected shall have a safe area for stopping a driver and must afford oncoming traffic sufficient sight distance for the driver to safely come to a stop upon approaching the checkpoint.
“2. The location must ensure minimum inconvenience for the driver and facilitate the safe stopping of traffic in one direction during the pilot program.
“3. Roadway choice must ensure that sufficient adjoining space is available to pull the vehicle off the traveled portion of the roadway for further inquiry if necessary.
“4. Consideration should be given to the physical space requirements as shown in Appendixes ‘A’ and ‘B.’” Id., at 149a-150a.
Although these criteria are not as open-ended as those used in Delaware v. Prouse, 440 U. S. 648 (1979), they certainly would permit the police to target an extremely wide variety of specific locations.
See, e. g., 1 Record 107.
It is, however, inappropriate for the Court to exaggerate that concern by relying on an outdated statistic from a tertiary source. The Court’s quotation from the 1987 edition of Professor LaFave’s treatise, ante, at 451, is in turn drawn from a 1983 law review note which quotes a 1982 House Committee Report that does not give the source for its figures. See 4 W. LaFave, Search and Seizure § 10.8(d), p. 71 (2d ed. 1987), citing Note, Curbing the Drunk Driver under the Fourth Amendment: The Constitutionality of Roadblock Seizures, 71 Geo. L. J. 1457, 1457, n. 1 (1983), citing, H. R. Rep. No. 97-867, p. 7.
Justice Blackmun’s citation, ante, at 455-456 (opinion concurring in judgment) to his own opinion in Perez v. Campbell, 402 U. S. 637, 657 (1971) (opinion concurring in part and dissenting in part) is even wider of the mark, since that case had nothing to do with drunken driving and the number of highway fatalities has since declined significantly despite the increase in highway usage.
By looking instead at recent data from the National Highway Traffic Safety Administration, one finds that in 1988 there were 18,501 traffic fatalities involving legally intoxicated persons and an additional 4,850 traffic fatalities involving persons with some alcohol exposure. Of course, the *468latter category of persons could not be arrested at a sobriety checkpoint, but even the total number of alcohol-related traffic fatalities (23,351) is significantly below the figure located by the student commentator and embraced by today’s Court. These numbers, of course, include any accidents that might have been caused by a sober driver but involved an intoxicated person. They also include accidents in which legally intoxicated pedestrians and bicyclists were killed; such accidents account for 2,180 of the 18,501 total accidents involving legally intoxicated persons. The checkpoints would presumably do nothing to intercept tipsy pedestrians or cyclists. See Fatal Accident Reporting System 1988 Overview, p. 1; id., Ch. 2, p. 5; see also 1 Record 58.
The dissents in those cases touted the relevant state interests in detail. In Almeida-Sanchez v. United States, 413 U. S. 266, 293 (1973), Justice White, joined by the author of today’s majority opinion, wrote:
“The fact is that illegal crossings at other than the legal ports of entry are numerous and recurring. If there is to be any hope of intercepting illegal entrants and of maintaining any kind of credible deterrent, it is essential that permanent or temporary checkpoints be maintained away from the borders, and roving patrols be conducted to discover and intercept illegal entrants as they filter to the established roads and highways and attempt to move away from the border area. It is for this purpose that the Border Patrol maintained the roving patrol involved in this case and conducted random, spot checks of automobiles and other vehicular traffic.”
Then-Justice Rehnquist argued in a similar vein in his dissent in Delaware v. Prouse, in which he observed that:
“The whole point of enforcing motor vehicle safety regulations is to remove from the road the unlicensed driver before he demonstrates why he is unlicensed.” 440 U. S., at 666.
“It is well to recall the words of Mr. Justice Jackson, soon after his return from the Nuremberg Trials:
“‘These [Fourth Amendment rights], I protest, are not mere second-class rights but belong in the catalog of indispensable freedoms. Among deprivations of rights, none is so effective in cowing a population, crushing the spirit of the individual and putting terror in every heart. Uncontrolled search and seizure is one of the first and most effective weapons in the arsenal of every arbitrary government.’ Brinegar v. United States, 338 U. S. 160, 180 [(1949)] (Jackson, J., dissenting).” Almeida-Sanchez v. United States, 413 U. S., at 273-274.
See, e. g., Walters v. National Assn. of Radiation Survivors, 473 U. S. 305, 371-372 (1985) (dissenting opinion); Hudson v. Palmer, 468 U. S. 517, 556-558 (1984) (dissenting opinion); Meachum v. Fano, 427 U. S. 215, 229-230 (1976) (dissenting opinion).
The Court refers to an expert’s testimony that the arrest rate is “around 1 percent,” ante, at 455, but a fair reading of the entire testimony of that witness, together with the other statistical evidence in the record, points to a significantly lower percentage.
Indeed, a single officer in a patrol car parked at the same place as the sobriety checkpoint would no doubt have been able to make some of the arrests based on the officer’s observation of the way the intoxicated driver was operating his vehicle.
Colonel Hough’s testimony included the following exchanges:
“Q. It is true, is it not, Colonel that your purpose in effectuating or attempting to effectuate this Checkpoint Plan is not to obtain large numbers of arrest of drunk drivers?
“A. That is correct.
“Q. Is it correct, is it, as far as you are aware, other states that have tried this have not found they are getting a high rate of arrests?
“A. Yes, that’s my understanding.
“Q. What was your purpose then, Colonel, in attempting to implement this plan if you don’t intend to use it to get drunk drivers arrested?
“A. Deter them from drinking and driving.” App. 77a.
“Q. To your knowledge, in the Maryland study, the part you reviewed, the check lanes are not an effective tool for arresting drunk drivers?
“A. They have not relied upon the number of arrests to judge the successfulness in my understanding.” Id., at 82a.
“Q. Are you aware that within the announcements that went out to the public was an indication that the checkpoints were to effectuate or [sic] arrest of drunk drivers. There was a goal to effectuate arrests of drunk drivers?
“A. Well, it is part of the role, sure.
“Q. Certainly not your primary goal, is it?
“A. The primary goal is to reduce alcohol related accidents.
“Q. It’s not your primary goal by any stretch, is it, to effectuate a high rate of arrests within this program?
*471“A. No.
“Q. If your goal was to effectuate a rise of arrests, you would use a different technique, wouldn’t you?
“A. I don’t know that.” 1 Record 88-89.
Respondents informed this Court that at trial “the Defendants did not even attempt to justify sobriety roadblocks on the basis of the number of arrests obtained.” Brief for Respondents 25. In answer, the State said: “Deterrence and public information are the primary goals of the sobriety checkpoint program, but the program is also clearly designed to apprehend any drunk drivers who pass through the checkpoint.” Reply Brief for Petitioner 34. This claim, however, does not directly controvert respondents’ argument or Colonel Hough’s concession: Even if the checkpoint is designed to produce some arrests, it does not follow that it has been adopted in order to produce arrests, or that it can be justified on such grounds.
“Dr. Ross’ testimony regarding the low actual arrest rate of checkpoint programs was corroborated by the testimony of one of defendants’ witnesses, Lieutenant Raymond Cotten of the Maryland State Police.” 170 Mich. App., at 442, 429 N. W. 2d, at 184.
“Our previous cases have recognized that maintenance of a traffic-checking program in the interior is necessary because the flow of illegal aliens cannot be controlled effectively at the border. We note here only the substantiality of the public interest in the practice of routine stops for inquiry at permanent checkpoints, a practice which the Government identi*472fies as the most important of the traffic-checking operations. Brief for United States in No. 74-1560, pp. 19-20. These checkpoints are located on important highways; in their absence such highways would offer illegal aliens a quick and safe route into the interior. Routine checkpoint inquiries apprehend many smugglers and illegal aliens who succumb to the lure of such highways. And the prospect of such inquiries forces others onto less efficient roads that are less heavily traveled, slowing their movement and making them more vulnerable to detection by roving patrols. Cf. United States v. Brignoni-Ponce, 422 U. S., at 883-885.
“A requirement that stops on major routes inland always be based on reasonable suspicion would be impractical because the flow of traffic tends to be too heavy to allow the particularized study of a given car that would enable it to be identified as a possible carrier of illegal aliens. In particular, such a requirement would largely eliminate any deterrent to the conduct of well-disguised smuggling operations, even though smugglers are known to use these highways regularly.” 428 U. S., at 556-557 (footnote omitted).
Alcohol-related traffic fatalities are also susceptible to reduction by public information campaigns in a way that crimes such as, for example, smuggling or armed assault are not. An intoxicated driver is her own most likely victim: More than 55 percent of those killed in accidents involving legally intoxicated drivers are legally intoxicated drivers themselves. Fatal Accident Reporting System 1988 Overview, p. 1. Cf. Skinner v. Railway Labor Executives’ Assn., 489 U. S. 602, 634 (Stevens, J., concurring in part and concurring in judgment) (“[I]f they are conscious of the possibilities that such an accident might occur and that alcohol or drug use might be a contributing factor, if the risk of serious personal injury does not deter their use of these substances, it seems highly unlikely that the *473additional threat of loss of employment would have any effect on their behavior”).
For example, in 1988 there were 18,501 traffic fatalities involving legally intoxicated persons. If one subtracts from this number the 10,210 legally intoxicated drivers who were themselves killed in these crashes, there remain 8,291 fatalities in which somebody other than the intoxicated driver was killed in an accident involving legally intoxicated persons (this *474number still includes, however, accidents in which legally intoxicated pedestrians stepped in front of sober drivers and were killed). Fatal Accident Reporting System 1988 Overview, p. 1; see also n. 15, supra.
By contrast, in 1986 there were a total of 19,257 murders and non-negligent manslaughters. Of these, approximately 11,360 were committed with a firearm, and another 3,850 were committed with some sort of knife. U. S. Dept. of Justice, 1987 Sourcebook of Criminal Justice Statistics 337 (1988).
From these statistics, it would seem to follow that someone who does not herself drive when legally intoxicated is more likely to be killed by an armed assailant than by an intoxicated driver. The threat to life from concealed weapons thus appears comparable to the threat from drunken driving.
Permanent, nondiscretionary checkpoints are already a common practice at public libraries, which now often require every patron to submit to a brief search for books, or to leave by passing through a special detector.
2 Record 40. Colonel Hough and Lieutenant Cotten agreed that publicity from the news media was an integral part of the checkpoint program. Colonel Hough, for example, testified as follows:
“Q. And you have observed, haven’t you, Colonel, any time you have a media campaign with regard to a crackdown you’re implementing, it does have a positive effect?
“A. We believe it has an effect, yes.
“Q. And in order for the positive effect of the media campaign to continue would be necessary to continue the announcements that you are putting out there?
“A. Yes.
“Q. It’s true, isn’t it, much of the media publicity attendant to this sobriety checkpoint has come from your public service announcements about the general media attention to this issue and placing it in our newspapers as a public interest story?
“A. Yes_
“Q. Or other television public interest stories?
“A. Yes.
“Q. You don’t anticipate, do you, Colonel, that the level of media interest in this matter will continue over the long haul, do you?
“A. I am certain it will wane in a period of time.
“Q. Have you ever given any thought to whether or not a different type of deterrent program with the same type of attendant media attention would have a similar deterrent effect as to what you can expect at the checkpoint?
“A. We have done it both with a SAVE Program and CARE Program and selective enforcement. Probably it has not received as great of attention as this has.
“Q. Any question, have you ever given any thought to whether or not a different technique with the same attendant media publicity that this has gotten would have the same effect you’re looking for here?
*476“A. No.” 1 id., at 91-92.
In addition, Point 6 of the Michigan State Police Sobriety Checkpoint Guidelines indicates that each driver stopped should be given a brochure describing the checkpoint’s purposes and operation. “The brochure will explain the purpose of the sobriety checkpoint program, furnish information concerning the effects of alcohol and safe consumption levels, and include a detachable pre-addressed questionnaire.” Trial Exhibit A, Michigan State Police Sobriety Checkpoint Guidelines 8 (Feb. 1986). The Maryland program had a similar feature. 2 Record 18.