Opinion

JUSTICE O'CONNOR delivered the opinion of the Court.

In Michigan Dept. of State Police v. Sitz, 496 U. S. 444 (1990), and United States v. Martinez-Fuerte, 428 U. S. 543 (1976), we held that brief, suspicionless seizures at highway checkpoints for the purposes of combating drunk driving and intercepting illegal immigrants were constitutional. We now consider the constitutionality of a highway checkpoint program whose primary purpose is the discovery and interdiction of illegal narcotics.

I

In August 1998, the city of Indianapolis began to operate vehicle checkpoints on Indianapolis roads in an effort to interdict unlawful drugs. The city conducted six such roadblocks between August and November that year, stopping

*Briefs of amici curiae urging reversal were filed for the State of Kansas et al. by Carla J. Stovall, Attorney General of Kansas, Stephen R. McAllister, State Solicitor, Jared S. Maag, Assistant Attorney General, and John M. Bailey, Chief State's Attorney of Connecticut, and by the Attorneys General for their respective States as follows: Bill Pryor of Alabama, Janet Napolitano of Arizona, Mark Pryor of Arkansas, Bill Lockyer of California, Robert A. Butterworth of Florida, James E. Ryan of Illinois, Karen M. Freeman-Wilson of Indiana, Thomas J. Miller of Iowa, Michael C. Moore of Mississippi, Don Stenberg of Nebraska, W A. Drew Edmondson of Oklahoma, Jan Graham of Utah, and Mark L. Earley of Virginia; for the National League of Cities et al. by Richard Ruda and James I. Crowley; and for the Washington Legal Foundation et al. by Daniel J. Popeo.

Briefs of amici curiae urging affirmance were filed for the National Association of Criminal Defense Lawyers et al. by Wesley MacNeil Oliver and Barbara Bergman; and for the Rutherford Institute by John W Whitehead and Steven H. Aden.

Wayne W Schmidt, James P. Manak, Richard Weintraub, and Bernard J. Farber filed a brief for Americans for Effective Law Enforcement, Inc., et al. as amici curiae. 1,161 vehicles and arresting 104 motorists. Fifty-five arrests were for drug-related crimes, while 49 were for offenses unrelated to drugs. Edmond v. Goldsmith, 183 F.3d 659, 661 (CA7 1999). The overall "hit rate" of the program was thus approximately nine percent.

The parties stipulated to the facts concerning the operation of the checkpoints by the Indianapolis Police Department (IPD) for purposes of the preliminary injunction proceedings instituted below. At each checkpoint location, the police stop a predetermined number of vehicles. Approximately 30 officers are stationed at the checkpoint. Pursuant to written directives issued by the chief of police, at least one officer approaches the vehicle, advises the driver that he or she is being stopped briefly at a drug checkpoint, and asks the driver to produce a license and registration. The officer also looks for signs of impairment and conducts an open-view examination of the vehicle from the outside. A narcoticsdetection dog walks around the outside of each stopped vehicle.

The directives instruct the officers that they may conduct a search only by consent or based on the appropriate quantum of particularized suspicion. The officers must conduct each stop in the same manner until particularized suspicion develops, and the officers have no discretion to stop any vehicle out of sequence. The city agreed in the stipulation to operate the checkpoints in such a way as to ensure that the total duration of each stop, absent reasonable suspicion or probable cause, would be five minutes or less.

The affidavit of Indianapolis Police Sergeant Marshall DePew, although it is technically outside the parties' stipulation, provides further insight concerning the operation of the checkpoints. According to Sergeant DePew, checkpoint locations are selected weeks in advance based on such considerations as area crime statistics and traffic flow. The checkpoints are generally operated during daylight hours and are identified with lighted signs reading, "'NARCOTICS CHECKPOINT __ MILE AHEAD, NARCOTICS K-9 IN USE, BE PREPARED TO STOP.'" App. to Pet. for Cert. 57a. Once a group of cars has been stopped, other traffic proceeds without interruption until all the stopped cars have been processed or diverted for further processing. Sergeant DePew also stated that the average stop for a vehicle not subject to further processing lasts two to three minutes or less.

Respondents James Edmond and J oell Palmer were each stopped at a narcotics checkpoint in late September 1998. Respondents then filed a lawsuit on behalf of themselves and the class of all motorists who had been stopped or were subject to being stopped in the future at the Indianapolis drug checkpoints. Respondents claimed that the roadblocks violated the Fourth Amendment of the United States Constitution and the search and seizure provision of the Indiana Constitution. Respondents requested declaratory and injunctive relief for the class, as well as damages and attorney's fees for themselves.

Respondents then moved for a preliminary injunction.

Although respondents alleged that the officers who stopped them did not follow the written directives, they agreed to the stipulation concerning the operation of the checkpoints for purposes of the preliminary injunction proceedings. The parties also stipulated to certification of the plaintiff class. The United States District Court for the Southern District of Indiana agreed to class certification and denied the motion for a preliminary injunction, holding that the checkpoint program did not violate the Fourth Amendment. Edmond v. Goldsmith, 38 F. Supp. 2d 1016 (1998). A divided panel of the United States Court of Appeals for the Seventh Circuit reversed, holding that the checkpoints contravened the Fourth Amendment. 183 F.3d 659 (1999). The panel denied rehearing. We granted certiorari, 528 U. S. 1153 (2000), and now affirm. II

The Fourth Amendment requires that searches and seizures be reasonable. A search or seizure is ordinarily unreasonable in the absence of individualized suspicion of wrongdoing. Chandler v. Miller, 520 U. S. 305, 308 (1997). While such suspicion is not an "irreducible" component of reasonableness, Martinez-Fuerte, 428 U. S., at 561, we have recognized only limited circumstances in which the usual rule does not apply. For example, we have upheld certain regimes of suspicionless searches where the program was designed to serve "special needs, beyond the normal need for law enforcement." See, e. g., Vernonia School Dist. J,7J v. Acton, 515 U. S. 646 (1995) (random drug testing of studentathletes); Treasury Employees v. Von Raab, 489 U. S. 656 (1989) (drug tests for United States Customs Service employees seeking transfer or promotion to certain positions); Skinner v. Railway Labor Executives' Assn., 489 U. S. 602 (1989) (drug and alcohol tests for railway employees involved in train accidents or found to be in violation of particular safety regulations). We have also allowed searches for certain administrative purposes without particularized suspicion of misconduct, provided that those searches are appropriately limited. See, e. g., New York v. Burger, 482 U. S. 691, 702-704 (1987) (warrantless administrative inspection of premises of "closely regulated" business); Michigan v. Tyler, 436 U. S. 499, 507-509, 511-512 (1978) (administrative inspection of fire-damaged premises to determine cause of blaze); Camara v. Municipal Court of City and County of San Francisco, 387 U. S. 523, 534-539 (1967) (administrative inspection to ensure compliance with city housing code).

We have also upheld brief, suspicionless seizures of motorists at a fixed Border Patrol checkpoint designed to intercept illegal aliens, Martinez-Fuerte, supra, and at a sobriety checkpoint aimed at removing drunk drivers from the road, Michigan Dept. of State Police v. Sitz, 496 U. S. 444 (1990). In addition, in Delaware v. Prouse, 440 U. S. 648, 663 (1979), we suggested that a similar type of roadblock with the purpose of verifying drivers' licenses and vehicle registrations would be permissible. In none of these cases, however, did we indicate approval of a checkpoint program whose primary purpose was to detect evidence of ordinary criminal wrongdoing.

In Martinez-Fuerte, we entertained Fourth Amendment challenges to stops at two permanent immigration checkpoints located on major United States highways less than 100 miles from the Mexican border. We noted at the outset the particular context in which the constitutional question arose, describing in some detail the "formidable law enforcement problems" posed by the northbound tide of illegal entrants into the United States. 428 U. S., at 551-554. These problems had also been the focus of several earlier cases addressing the constitutionality of other Border Patrol traffic-checking operations. See United States v. Ortiz, 422 U. S. 891 (1975); United States v. Brignoni-Ponce, 422 U. S. 873 (1975); Almeida-Sanchez v. United States, 413 U. S. 266 (1973). In Martinez-Fuerte, we found that the balance tipped in favor of the Government's interests in policing the Nation's borders. 428 U. S., at 561-564. In so finding, we emphasized the difficulty of effectively containing illegal immigration at the border itself. Id., at 556. We also stressed the impracticality of the particularized study of a given car to discern whether it was transporting illegal aliens, as well as the relatively modest degree of intrusion entailed by the stops. Id., at 556-564.

Our subsequent cases have confirmed that considerations specifically related to the need to police the border were a significant factor in our Martinez-Fuerte decision. For example, in United States v. Montoya de Hernandez, 473 U. S. 531,538 (1985), we counted Martinez-Fuerte as one of a number of Fourth Amendment cases that "reflect longstanding concern for the protection of the integrity of the border." Although the stops in Martinez-Fuerte did not occur at the border itself, the checkpoints were located near the border and served a border control function made necessary by the difficulty of guarding the border's entire length. See Martinez-Fuerte, supra, at 556.

In Sitz, we evaluated the constitutionality of a Michigan highway sobriety checkpoint program. The Sitz checkpoint involved brief, suspicionless stops of motorists so that police officers could detect signs of intoxication and remove impaired drivers from the road. 496 U. S., at 447-448. Motorists who exhibited signs of intoxication were diverted for a license and registration check and, if warranted, further sobriety tests. Id., at 447. This checkpoint program was clearly aimed at reducing the immediate hazard posed by the presence of drunk drivers on the highways, and there was an obvious connection between the imperative of highway safety and the law enforcement practice at issue. The gravity of the drunk driving problem and the magnitude of the State's interest in getting drunk drivers off the road weighed heavily in our determination that the program was constitutional. See id., at 451.

In Prouse, we invalidated a discretionary, suspicionless stop for a spot check of a motorist's driver's license and vehicle registration. The officer's conduct in that case was unconstitutional primarily on account of his exercise of "standardless and unconstrained discretion." 440 U. S., at 661. We nonetheless acknowledged the States' "vital interest in ensuring that only those qualified to do so are permitted to operate motor vehicles, that these vehicles are fit for safe operation, and hence that licensing, registration, and vehicle inspection requirements are being observed." Id., at 658. Accordingly, we suggested that "[q]uestioning of all oncoming traffic at roadblock-type stops" would be a lawful means of serving this interest in highway safety. Id., at 663.

We further indicated in Prouse that we considered the purposes of such a hypothetical roadblock to be distinct from a general purpose of investigating crime. The State prof fered the additional interests of "the apprehension of stolen motor vehicles and of drivers under the influence of alcohol or narcotics" in its effort to justify the discretionary spot check. Id., at 659, n. 18. We attributed the entirety of the latter interest to the State's interest in roadway safety. Ibid. We also noted that the interest in apprehending stolen vehicles may be partly subsumed by the interest in roadway safety. Ibid. We observed, however, that "[t]he remaining governmental interest in controlling automobile thefts is not distinguishable from the general interest in crime control." Ibid. Not only does the common thread of highway safety thus run through Sitz and Prouse, but Prouse itself reveals a difference in the Fourth Amendment significance of highway safety interests and the general interest in crime control.

III

It is well established that a vehicle stop at a highway checkpoint effectuates a seizure within the meaning of the Fourth Amendment. See, e. g., Sitz, supra, at 450. The fact that officers walk a narcotics-detection dog around the exterior of each car at the Indianapolis checkpoints does not transform the seizure into a search. See United States v. Place, 462 U. S. 696, 707 (1983). Just as in Place, an exterior sniff of an automobile does not require entry into the car and is not designed to disclose any information other than the presence or absence of narcotics. See ibid. Like the dog sniff in Place, a sniff by a dog that simply walks around a car is "much less intrusive than a typical search." Ibid. Cf. United States v. Turpin, 920 F.2d 1377, 1385 (CA8 1990). Rather, what principally distinguishes these checkpoints from those we have previously approved is their primary purpose.

As petitioners concede, the Indianapolis checkpoint program unquestionably has the primary purpose of interdicting illegal narcotics. In their stipulation of facts, the parties repeatedly refer to the checkpoints as "drug checkpoints" and describe them as "being operated by the City of Indianapolis in an effort to interdict unlawful drugs in Indianapolis." App. to Pet. for Cert. 51a-52a. In addition, the first document attached to the parties' stipulation is entitled "DRUG CHECKPOINT CONTACT OFFICER DIRECTIVES BY ORDER OF THE CHIEF OF POLICE." Id., at 53a. These directives instruct officers to "[a]dvise the citizen that they are being stopped briefly at a drug checkpoint." Ibid. The second document attached to the stipulation is entitled "1998 Drug Road Blocks" and contains a statistical breakdown of information relating to the checkpoints conducted. Id., at 55a. Further, according to Sergeant DePew, the checkpoints are identified with lighted signs reading, "'NARCOTICS CHECKPOINT __ MILE AHEAD, NARCOTICS K-9 IN USE, BE PREPARED TO STOP.'" Id., at 57a. Finally, both the District Court and the Court of Appeals recognized that the primary purpose of the roadblocks is the interdiction of narcotics. 38 F. Supp. 2d, at 1026 (noting that both parties "stress the primary purpose of the roadblocks as the interdiction of narcotics" and that "[t]he IPD has made it clear that the purpose for its checkpoints is to interdict narcotics traffic"); 183 F. 3d, at 665 (observing that "the City concedes that its proximate goal is to catch drug offenders").

We have never approved a checkpoint program whose primary purpose was to detect evidence of ordinary criminal wrongdoing. Rather, our checkpoint cases have recognized only limited exceptions to the general rule that a seizure must be accompanied by some measure of individualized suspicion. We suggested in Prouse that we would not credit the "general interest in crime control" as justification for a regime of suspicionless stops. 440 U. S., at 659, n. 18. Consistent with this suggestion, each of the checkpoint programs that we have approved was designed primarily to serve purposes closely related to the problems of policing the border or the necessity of ensuring roadway safety. Because the primary purpose of the Indianapolis narcotics checkpoint program is to uncover evidence of ordinary criminal wrongdoing, the program contravenes the Fourth Amendment.

Petitioners propose several ways in which the narcoticsdetection purpose of the instant checkpoint program may instead resemble the primary purposes of the checkpoints in Sitz and Martinez-Fuerte. Petitioners state that the checkpoints in those cases had the same ultimate purpose of arresting those suspected of committing crimes. Brief for Petitioners 22. Securing the border and apprehending drunk drivers are, of course, law enforcement activities, and law enforcement officers employ arrests and criminal prosecutions in pursuit of these goals. See Sitz, 496 U. S., at 447, 450; Martinez-Fuerte, 428 U. S., at 545-550. If we were to rest the case at this high level of generality, there would be little check on the ability of the authorities to construct roadblocks for almost any conceivable law enforcement purpose. Without drawing the line at roadblocks designed primarily to serve the general interest in crime control, the Fourth Amendment would do little to prevent such intrusions from becoming a routine part of American life.

Petitioners also emphasize the severe and intractable nature of the drug problem as justification for the checkpoint program. Brief for Petitioners 14-17, 31. There is no doubt that traffic in illegal narcotics creates social harms of the first magnitude. Cf. Von Raab, 489 U. S., at 668. The law enforcement problems that the drug trade creates likewise remain daunting and complex, particularly in light of the myriad forms of spin-off crime that it spawns. Cf. M on toya de Hernandez, 473 U. S., at 538. The same can be said of various other illegal activities, if only to a lesser degree. But the gravity of the threat alone cannot be dispositive of questions concerning what means law enforcement officers may employ to pursue a given purpose. Rather, in determining whether individualized suspicion is required, we must consider the nature of the interests threatened and their con nection to the particular law enforcement practices at issue. We are particularly reluctant to recognize exceptions to the general rule of individualized suspicion where governmental authorities primarily pursue their general crime control ends.

Nor can the narcotics-interdiction purpose of the checkpoints be rationalized in terms of a highway safety concern similar to that present in Sitz. The detection and punishment of almost any criminal offense serves broadly the safety of the community, and our streets would no doubt be safer but for the scourge of illegal drugs. Only with respect to a smaller class of offenses, however, is society confronted with the type of immediate, vehicle-bound threat to life and limb that the sobriety checkpoint in Sitz was designed to eliminate.

Petitioners also liken the anticontraband agenda of the Indianapolis checkpoints to the antismuggling purpose of the checkpoints in Martinez-Fuerte. Brief for Petitioners 1516. Petitioners cite this Court's conclusion in MartinezFuerte that the flow of traffic was too heavy to permit "particularized study of a given car that would enable it to be identified as a possible carrier of illegal aliens," 428 U. S., at 557, and claim that this logic has even more force here. The problem with this argument is that the same logic prevails any time a vehicle is employed to conceal contraband or other evidence of a crime. This type of connection to the roadway is very different from the close connection to roadway safety that was present in Sitz and Prouse. Further, the Indianapolis checkpoints are far removed from the border context that was crucial in Martinez-Fuerte. While the difficulty of examining each passing car was an important factor in validating the law enforcement technique employed in Martinez-Fuerte, this factor alone cannot justify a regime of suspicionless searches or seizures. Rather, we must look more closely at the nature of the public interests that such a regime is designed principally to serve. The primary purpose of the Indianapolis narcotics checkpoints is in the end to advance "the general interest in crime control," Prouse, 440 U. S., at 659, n. 18. We decline to suspend the usual requirement of individualized suspicion where the police seek to employ a checkpoint primarily for the ordinary enterprise of investigating crimes. We cannot sanction stops justified only by the generalized and everpresent possibility that interrogation and inspection may reveal that any given motorist has committed some crime.

Of course, there are circumstances that may justify a law enforcement checkpoint where the primary purpose would otherwise, but for some emergency, relate to ordinary crime control. For example, as the Court of Appeals noted, the Fourth Amendment would almost certainly permit an appropriately tailored roadblock set up to thwart an imminent terrorist attack or to catch a dangerous criminal who is likely to flee by way of a particular route. See 183 F. 3d, at 662663. The exigencies created by these scenarios are far removed from the circumstances under which authorities might simply stop cars as a matter of course to see if there just happens to be a felon leaving the jurisdiction. While we do not limit the purposes that may justify a checkpoint program to any rigid set of categories, we decline to approve a program whose primary purpose is ultimately indistinguishable from the general interest in crime control.l

THE CHIEF JUSTICE'S dissent also erroneously characterizes our opinion as holding that the "use of a drug-sniffing dog... annuls what is otherwise plainly constitutional under our Fourth Amendment jurisprudence." Post, at 48. Again, the constitutional defect of the program is that its primary purpose is to advance the general interest in crime control. Petitioners argue that our prior cases preclude an inquiry into the purposes of the checkpoint program. For example, they cite Whren v. United States, 517 U. S. 806 (1996), and Bond v. United States, 529 U. S. 334 (2000), to support the proposition that "where the government articulates and pursues a legitimate interest for a suspicionless stop, courts should not look behind that interest to determine whether the government's 'primary purpose' is valid." Brief for Petitioners 34; see also id., at 9. These cases, however, do not control the instant situation.

In Whren, we held that an individual officer's subjective intentions are irrelevant to the Fourth Amendment validity of a traffic stop that is justified objectively by probable cause to believe that a traffic violation has occurred. 517 U. S., at 810-813. We observed that our prior cases "foreclose any argument that the constitutional reasonableness of traffic stops depends on the actual motivations of the individual officers involved." Id., at 813. In so holding, we expressly distinguished cases where we had addressed the validity of searches conducted in the absence of probable cause. See id., at 811-812 (distinguishing Florida v. Wells, 495 U. S. 1, 4 (1990) (stating that "an inventory search must not be a ruse for a general rummaging in order to discover incriminating evidence"), Colorado v. Bertine, 479 U. S. 367, 372 (1987) (suggesting that the absence of bad faith and the lack of a purely investigative purpose were relevant to the validity of an inventory search), and Burger, 482 U. S., at 716-717, n. 27 (observing that a valid administrative inspection conducted with neither a warrant nor probable cause did not appear to be a pretext for gathering evidence of violations of the penal laws)).

Whren therefore reinforces the principle that, while "[sJubjective intentions play no role in ordinary, probablecause Fourth Amendment analysis," 517 U. S., at 813, programmatic purposes may be relevant to the validity of Fourth Amendment intrusions undertaken pursuant to a general scheme without individualized SuspICIOn. Accordingly, Whren does not preclude an inquiry into programmatic purpose in such contexts. Cf. Chandler v. Miller, 520 U. S. 305 (1997); Treasury Employees v. Von Raab, 489 U. S. 656 (1989); Burger, supra; Michigan v. Tyler, 436 U. S. 499 (1978); Camara v. Municipal Court of City and County of San Francisco, 387 U. S. 523 (1967). It likewise does not preclude an inquiry into programmatic purpose here.

Last Term in Bond, we addressed the question whether a law enforcement officer violated a reasonable expectation of privacy in conducting a tactile examination of carry-on luggage in the overhead compartment of a bus. In doing so, we simply noted that the principle of Whren rendered the subjective intent of an officer irrelevant to this analysis. 529 U. S., at 338, n. 2. While, as petitioners correctly observe, the analytical rubric of Bond was not "ordinary, probable-cause Fourth Amendment analysis," Whren, supra, at 813, nothing in Bond suggests that we would extend the principle of Whren to all situations where individualized suspicion was lacking. Rather, subjective intent was irrelevant in Bond because the inquiry that our precedents required focused on the objective effects of the actions of an individual officer. By contrast, our cases dealing with intrusions that occur pursuant to a general scheme absent individualized suspicion have often required an inquiry into purpose at the programmatic level.

Petitioners argue that the Indianapolis checkpoint program is justified by its lawful secondary purposes of keeping impaired motorists off the road and verifying licenses and registrations. Brief for Petitioners 31-34. If this were the case, however, law enforcement authorities would be able to establish checkpoints for virtually any purpose so long as they also included a license or sobriety check. For this reason, we examine the available evidence to determine the primary purpose of the checkpoint program. While we recognize the challenges inherent in a purpose inquiry, courts routinely engage in this enterprise in many areas of constitutional jurisprudence as a means of sifting abusive governmental conduct from that which is lawful. Cf. 183 F. 3d, at 665. As a result, a program driven by an impermissible purpose may be proscribed while a program impelled by licit purposes is permitted, even though the challenged conduct may be outwardly similar. While reasonableness under the Fourth Amendment is predominantly an objective inquiry, our special needs and administrative search cases demonstrate that purpose is often relevant when suspicionless intrusions pursuant to a general scheme are at issue.2

It goes without saying that our holding today does nothing to alter the constitutional status of the sobriety and border checkpoints that we approved in Sitz and Martinez-Fuerte, or of the type of traffic checkpoint that we suggested would be lawful in Prouse. The constitutionality of such checkpoint programs still depends on a balancing of the competing interests at stake and the effectiveness of the program. See Sitz, 496 U. S., at 450-455; Martinez-Fuerte, 428 U. S., at 556-564. When law enforcement authorities pursue primarily general crime control purposes at checkpoints such as here, however, stops can only be justified by some quantum of individualized suspicion.

Our holding also does not affect the validity of border searches or searches at places like airports and government buildings, where the need for such measures to ensure public safety can be particularly acute. Nor does our opinion speak to other intrusions aimed primarily at purposes beyond the general interest in crime control. Our holding also does not impair the ability of police officers to act appropriately upon information that they properly learn during a checkpoint stop justified by a lawful primary purpose, even where such action may result in the arrest of a motorist for an offense unrelated to that purpose. Finally, we caution that the purpose inquiry in this context is to be conducted only at the programmatic level and is not an invitation to probe the minds of individual officers acting at the scene. Cf. Whren, supra.

Because the primary purpose of the Indianapolis checkpoint program is ultimately indistinguishable from the general interest in crime control, the checkpoints violate the Fourth Amendment. The judgment of the Court of Appeals is, accordingly, affirmed.

It is so ordered.

Notes

1 THE CHIEF JUSTICE'S dissent erroneously characterizes our opinion as resting on the application of a "non-law-enforcement primary purpose test." Post, at 53. Our opinion nowhere describes the purposes of the Sitz and Martinez-Fuerte checkpoints as being "not primarily related to criminal law enforcement." Post, at 50. Rather, our judgment turns on the fact that the primary purpose of the Indianapolis checkpoints is to advance the general interest in crime control.

2 Because petitioners concede that the primary purpose of the Indianapolis checkpoints is narcotics detection, we need not decide whether the State may establish a checkpoint program with the primary purpose of checking licenses or driver sobriety and a secondary purpose of interdicting narcotics. Specifically, we express no view on the question whether police may expand the scope of a license or sobriety checkpoint seizure in order to detect the presence of drugs in a stopped car. Cf. New Jersey v. T. L. Q, 469 U. S. 325, 341 (1985) (search must be "'reasonably related in scope to the circumstances which justified the interference in the first place'" (quoting Terry v. Ohio, 392 U. S. 1, 20 (1968))); Michigan v. Clif ford, 464 U. S. 287, 294-295 (1984) (plurality opinion).