with whom Justice O’Connor joins, dissenting.

Without expressing any views on the merits of this case, I respectfully dissent because in my opinion we have no authority to decide it. I think the controversy is moot.

I

The Court correctly acknowledges that we have no power under Art. Ill of the Constitution to adjudicate a case that no *333longer presents an actual, ongoing dispute between the named parties. Ante, at 317, citing Nebraska Press Assn. v. Stuart, 427 U. S. 539, 546 (1976); Preiser v. Newkirk, 422 U. S. 395, 401 (1975). Here, there is obviously no present controversy between the parties, since both respondents are no longer in school and therefore no longer subject to a unilateral “change in placement.” The Court concedes mootness with respect to respondent John Doe, who is now too old to receive the benefits of the Education of the Handicapped Act (EHA). Ante, at 318. It concludes, however, that the case is not moot as to respondent Jack Smith, who has two more years of eligibility but is no longer in the public schools, because the controversy is “capable of repetition, yet evading review.” Ante, at 318-323.

Jurisdiction on the basis that a dispute is “capable of repetition, yet evading review” is limited to the “exceptional situation],” Los Angeles v. Lyons, 461 U. S. 95, 109 (1983), where the following two circumstances simultaneously occur: “ ‘(1) the challenged action [is] in its duration too short to be fully litigated prior to its cessation or expiration, and (2) there [is] a reasonable expectation that the same complaining party would be subjected to the same action again.’” Murphy v. Hunt, 455 U. S. 478, 482 (1982) (per curiam), quoting Weinstein v. Bradford, 423 U. S. 147, 149 (1975) (per curiam). The second of these requirements is not met in this case.

For there to be a “reasonable expectation” that Smith will be subjected to the same action again, that event must be a “demonstrated probability.” Murphy v. Hunt, supra, at 482, 483; Weinstein v. Bradford, supra, at 149. I am surprised by the Court’s contention, fraught with potential for future mischief, that “reasonable expectation” is satisfied by something less than “demonstrated probability.” Ante, at 318-319, n. 6. No one expects that to happen which he does not think probable; and his expectation cannot be shown to be reasonable unless the probability is demonstrated. Thus, as the Court notes, our cases recite the two descriptions side by *334side (“a ‘reasonable expectation’ or a ‘demonstrated probability,’” Hunt, supra, at 482). The Court asserts, however, that these standards are “described ... in the disjunctive,” ante, at 318-319, n. 6 — evidently believing that the conjunction “or” has no accepted usage except a disjunctive one, i. e., “expressing an alternative, contrast, or opposition,” Webster’s Third New International Dictionary 651 (1981). In fact, however, the conjunction is often used “to indicate ... (3) the synonymous, equivalent, or substitutive character of two words or phrases (fell over a precipice [or] cliff) (the off [or] far side) (lessen [or] abate); (4) correction or greater exactness of phrasing or meaning (these essays, [or] rather rough sketches) (the present king had no children— [or] no legitimate children . . .).” Id., at 1585. It is obvious that in saying “a reasonable expectation or a demonstrated probability” we have used the conjunction in one of the latter, or nondisjunctive, senses. Otherwise (and according to the Court’s exegesis), we would have been saying that a controversy is sufficiently likely to recur if either a certain degree of probability exists or a higher degree of probability exists. That is rather like a statute giving the vote to persons who are “18 or 21.” A bare six years ago, the author of today’s opinion and one other Member of the majority plainly understood “reasonable expectation” and “demonstrated probability” to be synonymous. Cf. Edgar v. MITE Corp., 457 U. S. 624, 662, and n. 11 (1982) (Marshall, J., dissenting, joined by Brennan, J.) (using the two terms here at issue interchangeably, and concluding that the case is moot because “there is no demonstrated probability that the State will have occasion to prevent MITE from making a takeover offer for some other corporation”) (emphasis added).

The prior holdings cited by the Court in a footnote, see ante, at 319, n. 6, offer no support for the novel proposition that less than a probability of recurrence is sufficient to avoid mootness. In Burlington Northern R. Co. v. Maintenance of Way Employes, 481 U. S. 429, 436, n. 4 (1987), we found *335that the same railroad and union were “reasonably likely” to find themselves in a recurring dispute over the same issue. Similarly, in California Coastal Comm’n v. Granite Rock Co., 480 U. S. 572, 578 (1987), we found it “likely” that the plaintiff mining company would submit new plans which the State would seek to subject to its coastal permit requirements. See Webster’s Third New International Dictionary 1310 (1981) (defining “likely” as “of such a nature or so circumstanced as to make something probable[;]. . . seeming to justify belief or expectation^]... in all probability”). In the cases involving exclusion orders issued to prevent the press from attending criminal trials, we found that “[i]t can reasonably be assumed” that a news organization covering the area in which the defendant court sat will again be subjected to that court’s closure rules. Press-Enterprise Co. v. Superior Court of Cal., Riverside County, 478 U. S. 1, 6 (1986); Globe Newspaper Co. v. Superior Court of Norfolk County, 457 U. S. 596, 603 (1982). In these and other cases, one may quarrel, perhaps, with the accuracy of the Court’s probability assessment; but there is no doubt that assessment was regarded as necessary to establish jurisdiction.

In Roe v. Wade, 410 U. S. 113, 125 (1973), we found that the “human gestation period is so short that the pregnancy will come to term before the usual appellate process is complete,” so that “pregnancy litigation seldom will survive much beyond the trial stage, and appellate review will be effectively denied.” Roe, at least one other abortion case, see Doe v. Bolton, 410 U. S. 179, 187 (1973), and some of our election law decisions, see Rosario v. Rockefeller, 410 U. S. 752, 756, n. 5 (1973); Dunn v. Blumstein, 405 U. S. 330, 333, n. 2 (1972), differ from the body of our mootness jurisprudence not in accepting less than a probability that the issue will recur, in a manner evading review, between the same parties; but in dispensing with the same-party requirement entirely, focusing instead upon the great likelihood that the issue will recur between the defendant and the other members *336of the public at large without ever reaching us. Arguably those cases have been limited to their facts, or to the narrow areas of abortion and election rights, by our more recent insistence that, at least in the absence of a class action, the “capable of repetition” doctrine applies only where “there [is] a ‘reasonable expectation’” that the “same complaining party” would be subjected to the same action again. Hunt, 455 U. S., at 482 (emphasis added), quoting Weinstein, 423 U. S., at 149; see Burlington Northern R. Co., supra, at 436, n. 4; Illinois Elections Bd. v. Socialist Workers Party, 440 U. S. 173, 187 (1979). If those earlier cases have not been so limited, however, the conditions for their application do not in any event exist here. There is no extraordinary improbability of the present issue’s reaching us as a traditionally live controversy. It would have done so in this very case if Smith had not chosen to leave public school. In sum, on any analysis, the proposition the Court asserts in the present case — that probability need not be shown in order to establish the “same-party-recurrence” exception to mootness —is a significant departure from settled law.

II

If our established mode of analysis were followed, the conclusion that a live controversy exists in the present case would require a demonstrated probability that all of the following events will occur: (1) Smith will return to public school; (2) he will be placed in an educational setting that is unable to tolerate his dangerous behavior; (3) he will again engage in dangerous behavior; and (4) local school officials will again attempt unilaterally to change his placement and the state defendants will fail to prevent such action. The Court spends considerable time establishing that the last two of these events are likely to recur, but relegates to a footnote its discussion of the first event, upon which all others depend, and only briefly alludes to the second. Neither the facts in *337the record, nor even the extrarecord assurances of counsel, establish a demonstrated probability of either of them.

With respect to whether Smith will return to school, at oral argument Smith’s counsel forthrightly conceded that she “cannot represent whether in fact either of these students will ask for further education from the Petitioners.” Tr. of Oral Arg. 23. Rather, she observed, respondents would “look to [our decision in this case] to find out what will happen after that.” Id., at 23-24. When pressed, the most counsel would say was that, in her view, the 20-year-old Smith could seek to return to public school because he has not graduated, he is handicapped, and he has a right to an education. Id., at 27. I do not perceive the principle that would enable us to leap from the proposition that Smith could reenter public school to the conclusion that it is a demonstrated probability he will do so.

The Court nevertheless concludes that “there is at the very least a reasonable expectation” that Smith will return to school. Ante, at 319, n. 6. I cannot possibly dispute that on the basis of the Court’s terminology. Once it is accepted that a “reasonable expectation” can exist without a demonstrable probability that the event in question will occur, the phrase has been deprived of all meaning, and the Court can give it whatever application it wishes without fear of effective contradiction. It is worth pointing out, however, how slim are the reeds upon which this conclusion of “reasonable expectation” (whatever that means) rests. The Court bases its determination on three observations from the record and oral argument. First, it notes that Smith has been pressing this lawsuit since 1980. It suffices to observe that the equivalent argument can be made in every case that remains active and pending; we have hitherto avoided equating the existence of a case or controversy with the existence of a lawsuit. Second, the Court observes that Smith has “as great a need of a high school education and diploma as any of his peers.” Ibid. While this is undoubtedly good advice, it hardly estab*338lishes that the 20-year-old Smith is likely to return to high school, much less to public high school. Finally, the Court notes that counsel “advises us that [Smith] is awaiting the outcome of this case to decide whether to pursue his degree.” Ibid. Not only do I not think this establishes a current case or controversy, I think it a most conclusive indication that no current case or controversy exists. We do not sit to broaden decisionmaking options, but to adjudicate the lawfulness of acts that have happened or, at most, are about to occur.

The conclusion that the case is moot is reinforced, moreover, when one considers that, even if Smith does return to public school, the controversy will still not recur unless he is again placed in an educational setting that is unable to tolerate his behavior. It seems to me not only not demonstrably probable, but indeed quite unlikely, given what is now known about Smith’s behavioral problems, that local school authorities would again place him in an educational setting that could not control his dangerous conduct, causing a suspension that would replicate the legal issues in this suit. The majority dismisses this further contingency by noting that the school authorities have an obligation under the EHA to provide an “appropriate” education in “the least restrictive environment.” Ante, at 321. This means, however, the least restrictive environment appropriate for the particular child. The Court observes that “the preparation of an [individualized educational placement]” is “an inexact science at best,” ibid., thereby implying that the school authorities are likely to get it wrong. Even accepting this assumption, which seems to me contrary to the premises of the Act, I see no reason further to assume that they will get it wrong by making the same mistake they did last time — assigning Smith to too ^restrictive an environment, from which he will thereafter be suspended — rather than by assigning him to too restrictive an environment. The latter, which seems to me more likely than the former (though both combined are much less likely than a correct placement), might produce a law*339suit, but not a lawsuit involving the issues that we have before us here.

Ill

The Chief Justice joins the majority opinion on the ground, not that this case is not moot, but that where the events giving rise to the mootness have occurred after we have granted certiorari we may disregard them, since mootness is only a prudential doctrine and not part of the “case or controversy” requirement of Art. III. I do not see how that can be. There is no more reason to intuit that mootness is merely a prudential doctrine than to intuit that initial standing is. Both doctrines have equivalently deep roots in the common-law understanding, and hence the constitutional understanding, of what makes a matter appropriate for judicial disposition. See Flast v. Cohen, 392 U. S. 83, 95 (1968) (describing mootness and standing as various illustrations of the requirement of “justiciability” in Art. III).

The Chief Justice relies upon the fact that an 1895 case discussing mootness, Mills v. Green, 159 U. S. 651, makes no mention of the Constitution. But there is little doubt that the Court believed the doctrine called into question the Court’s power and not merely its prudence, for (in an opinion by the same Justice who wrote Mills) it had said two years earlier:

“[T]he court is not empowered to decide moot questions or abstract propositions, or to declare . . . principles or rules of law which cannot affect the result as to the thing in issue in the case before it. No stipulation of parties or counsel. . . can enlarge the power, or affect the duty, of the court in this regard.” California v. San Pablo & Tulare R. Co., 149 U. S. 308, 314 (1893) (Gray, J.) (emphasis added).

If it seems peculiar to the modern lawyer that our 19th-century mootness cases make no explicit mention of Art. Ill, that is a peculiarity shared with our 19th-century, and even *340our early 20th-century, standing cases. As late as 1919, in dismissing a suit for lack of standing we said simply:

“Considerations of propriety, as well as long-established practice, demand that we refrain from passing upon the constitutionality of an act of Congress unless obliged to do so in the proper performance of our judicial function, when the question is raised by a party whose interests entitle him to raise it.” Blair v. United States, 250 U. S. 273, 279.

See also, e. g., Standard Stock Food Co. v. Wright, 225 U. S. 540, 550 (1912); Southern R. Co. v. King, 217 U. S. 524, 534 (1910); Turpin v. Lemon, 187 U. S. 51, 60-61 (1902); Tyler v. Judges of Court of Registration, 179 U. S. 405, 409 (1900). The same is also true of our early cases dismissing actions lacking truly adverse parties, that is, collusive actions. See, e. g., Cleveland v. Chamberlain, 1 Black 419, 425-426 (1862); Lord v. Veazie, 8 How. 251, 254-256 (1850). The explanation for this ellipsis is that the courts simply chose to refer directly to the traditional, fundamental limitations upon the powers of common-law courts, rather than referring to Art. Ill which in turn adopts those limitations through terms (“The judicial Power”; “Cases”; “Controversies”) that have virtually no meaning except by reference to that tradition. The ultimate circularity, coming back in the end to tradition, is evident in the statement by Justice Field:

“By cases and controversies are intended the claims of litigants brought before the courts for determination by such regular proceedings as are established by law or custom for the protection or enforcement of rights, or the prevention, redress, or punishment of wrongs. Whenever the claim of a party under the constitution, laws, or treaties of the United States takes such a form that the judicial power is capable of acting upon it, then it has become a case.” In re Pacific Railway Comm’n, 32 F. 241, 255 (CC ND Cal. 1887).

*341See also 2 M. Farrand, Records of the Federal Convention of 1787, p. 430 (rev. ed. 1966):

“Doer. Johnson moved to insert the words ‘this Constitution and the’ before the word ‘laws.’

“Mr. Madison doubted whether it was not going too far to extend the jurisdiction of the Court generally to cases arising Under the Constitution, & whether it ought not to be limited to cases of a Judiciary Nature. The right of expounding the Constitution in cases not of this nature ought not to be given to that Department.

“The motion of Doer. Johnson was agreed to nem: con: it being generally supposed that the jurisdiction given was constructively limited to cases of a Judiciary nature.”

In sum, I cannot believe that it is only our prudence, and nothing inherent in the understood nature of “The judicial Power,” U. S. Const., Art. Ill, §1, that restrains us from pronouncing judgment in a case that the parties have settled, or a case involving a nonsurviving claim where the plaintiff has died, or a case where the law has been changed so that the basis of the dispute no longer exists, or a case where conduct sought to be enjoined has ceased and will not recur. Where the conduct has ceased for the time being but there is a demonstrated probability that it will recur, a real-life controversy between parties with a personal stake in the outcome continues to exist, and Art. Ill is no more violated than it is violated by entertaining a declaratory judgment action. But that is the limit of our power. I agree with The Chief Justice to this extent: the “yet evading review” portion of our “capable of repetition, yet evading review” test is prudential; whether or not that criterion is met, a justiciable controversy exists. But the probability of recurrence between the same parties is essential to our jurisdiction as a court, and it is that deficiency which the case before us presents.

It is assuredly frustrating to find that a jurisdictional impediment prevents us from reaching the important merits *342issues that were the reason for our agreeing to hear this case. But we cannot ignore such impediments for purposes of our appellate review without simultaneously affecting the principles that govern district courts in their assertion or retention of original jurisdiction. We thus do substantial harm to a governmental structure designed to restrict the courts to matters that actually affect the litigants before them.

delivered the opinion of the Court.

As a condition of federal financial assistance, the Education of the Handicapped Act requires States to ensure a “free appropriate public education” for all disabled children within their jurisdictions. In aid of this goal, the Act establishes a comprehensive system of procedural safeguards designed to ensure parental participation in decisions concerning the education of their disabled children and to provide administrative and judicial review of any decisions with which those parents disagree. Among these safeguards is the so-called “stay-put” provision, which directs that a disabled child “shall remain in [his or her] then current educational placement” pending completion of any review proceedings, unless the parents and state or local educational agencies otherwise agree. 20 U. S. C. § 1415(e)(3). Today we must decide whether, in the face of this statutory proscription, state or local school authorities may nevertheless unilaterally exclude disabled children from the classroom for dangerous or disruptive conduct growing out of their disabilities. In addition, we are called upon to decide whether a district court may, in the exercise of its equitable powers, order a State to provide educational services directly to a disabled child when the local agency fails to do so.

*309I

In the Education of the Handicapped Act (EHA or the Act), 84 Stat. 175, as amended, 20 U. S. C. §1400 et seq., Congress sought “to assure that all handicapped children have available to them ... a free appropriate public education which emphasizes special education and related services designed to meet their unique needs, [and] to assure that the rights of handicapped children and their parents or guardians are protected.” § 1400(c). When the law was passed in 1975, Congress had before it ample evidence that such legislative assurances were sorely needed: 21 years after this Court declared education to be “perhaps the most important function of state and local governments,” Brown v. Board of Education, 347 U. S. 483, 493 (1954), congressional studies revealed that better than half of the Nation’s 8 million disabled children were not receiving appropriate educational services. § 1400(b)(3). Indeed, one out of every eight of these children was excluded from the public school system altogether, § 1400(b)(4); many others were simply “warehoused” in special classes or were neglectfully shepherded through the system until they were old enough to drop out. See H. R. Rep. No. 94-332, p. 2 (1975). Among the most poorly served of disabled students were emotionally disturbed children: Congressional statistics revealed that for the school year immediately preceding passage of the Act, the educational needs of 82 percent of all children with emotional disabilities went unmet. See S. Rep. No. 94-168, p. 8 (1975) (hereinafter S. Rep.).

Although these educational failings resulted in part from funding constraints, Congress recognized that the problem reflected more than a lack of financial resources at the state and local levels. Two federal-court decisions, which the Senate Report characterized as “landmark,” see id., at 6, demonstrated that many disabled children were excluded pursuant to state statutes or local rules and policies, typically without *310any consultation with, or even notice to, their parents. See Mills v. Board of Education of District of Columbia, 348 F. Supp. 866 (DC 1972); Pennsylvania Assn. for Retarded Children v. Pennsylvania, 334 F. Supp. 1257 (ED Pa. 1971), and 343 F. Supp. 279 (1972) (PARC). Indeed, by the time of the EHA’s enactment, parents had brought legal challenges to similar exclusionary practices in 27 other States. See S. Rep., at 6.

In responding to these problems, Congress did not content itself with passage of a simple funding statute. Rather, the EHA confers upon disabled students an enforceable substantive right to public education in participating States, see Board of Education of Hendrick Hudson Central School Dist. v. Rowley, 458 U. S. 176 (1982),1 and conditions federal financial assistance upon a State’s compliance with the substantive and procedural goals of the Act. Accordingly, States seeking to qualify for federal funds must develop policies assuring all disabled children the “right to a free appropriate public education,” and must file with the Secretary of *311Education formal plans mapping out in detail the programs, procedures, and timetables under which they will effectuate these policies. 20 U. S. C. §§ 1412(1), 1413(a). Such plans must assure that, “to the maximum extent appropriate,” States will “mainstream” disabled children, i. e., that they will educate them with children who are not disabled, and that they will segregate or otherwise remove such children from the regular classroom setting “only when the nature or severity of the handicap is such that education in regular classes . . . cannot be achieved satisfactorily.” § 1412(5).

The primary vehicle for implementing these congressional goals is the “individualized educational program” (IEP), which the EHA mandates for each disabled child. Prepared at meetings between a representative of the local school district, the child’s teacher, the parents or guardians, and, whenever appropriate, the disabled child, the IEP sets out the child’s present educational performance, establishes annual and short-term objectives for improvements in that performance, and describes the specially designed instruction and services that will enable the child to meet those objectives. § 1401(19). The IEP must be reviewed and, where necessary, revised at least once a year in order to ensure that local agencies tailor the statutorily required “free appropriate public education” to each child’s unique needs. § 1414(a)(5).

Envisioning the IEP as the centerpiece of the statute’s education delivery system for disabled children, and aware that schools had all too often denied such children appropriate educations without in any way consulting their parents, Congress repeatedly emphasized throughout the Act the importance and indeed the necessity of parental participation in both the development of the IEP and any subsequent assessments of its effectiveness. See §§ 1400(c), 1401(19), 1412(7), 1415(b)(1)(A), (C), (D), (E), and 1415(b)(2). Accordingly, the Act establishes various procedural safeguards that guarantee parents both an opportunity for meaningful input into all decisions affecting their child’s education and the right *312to seek review of any decisions they think inappropriate. These safeguards include the right to examine all relevant records pertaining to the identification, evaluation, and educational placement of their child; prior written notice whenever the responsible educational agency proposes (or refuses) to change the child’s placement or program; an opportunity to present complaints concerning any aspect of the local agency’s provision of a free appropriate public education; and an opportunity for “an impartial due process hearing” with respect to any such complaints. §§ 1415(b)(1), (2).

At the conclusion of any such hearing, both the parents and the local educational agency may seek further administrative review and, where that proves unsatisfactory, may file a civil action in any state or federal court. §§ 1415(c), (e)(2). In addition to reviewing the administrative record, courts are empowered to take additional evidence at the request of either party and to “grant such relief as [they] determine[ ] is appropriate.” § 1415(e)(2). The “stay-put” provision at issue in this case governs the placement of a child while these often lengthy review procedures run their course. It directs that:

“During the pendency of any proceedings conducted pursuant to [§ 1415], unless the State or local educational agency and the parents or guardian otherwise agree, the child shall remain in the then current educational placement of such child . . . .” § 1415(e)(3).

The present dispute grows out of the efforts of certain officials of the San Francisco Unified School District (SFUSD) to expel two emotionally disturbed children from school indefinitely for violent and disruptive conduct related to their disabilities. In November 1980, respondent John Doe assaulted another student at the Louise Lombard School, a developmental center for disabled children. Doe’s April 1980 IEP identified him as a socially and physically awkward 17-year-old who experienced considerable difficulty controlling his impulses and anger. Among the goals set out in his IEP was “[i]mprovement in [his] ability to relate to [his] *313peers [and to] cope with frustrating situations without resorting to aggressive acts.” App. 17. Frustrating situations, however, were an unfortunately prominent feature of Doe’s school career: physical abnormalities, speech difficulties, and poor grooming habits had made him the target of teasing and ridicule as early as the first grade, id., at 23; his 1980 IEP reflected his continuing difficulties with peers, noting that his social skills had deteriorated and that he could tolerate only minor frustration before exploding. Id., at 15-16.

On November 6, 1980, Doe responded to the taunts of a fellow student in precisely the explosive manner anticipated by his IEP: he choked the student with sufficient force to leave abrasions on the child’s neck, and kicked out a school window while being escorted to the principal’s office afterwards. Id., at 208. Doe admitted his misconduct and the school subsequently suspended him for five days. Thereafter, his principal referred the matter to the SFUSD Student Placement Committee (SPC or Committee) with the recommendation that Doe be expelled. On the day the suspension was to end, the SPC notified Doe’s mother that it was proposing to exclude her child permanently from SFUSD and was therefore extending his suspension until such time as the expulsion proceedings were completed.2 The Committee further advised her that she was entitled to attend the November 25 hearing at which it planned to discuss the proposed expulsion.

After unsuccessfully protesting these actions by letter, Doe brought this suit against a host of local school officials *314and the State Superintendent of Public Instruction. Alleging that the suspension and proposed expulsion violated the EHA, he sought a temporary restraining order canceling the SPC hearing and requiring school officials to convene an IEP meeting. The District Judge granted the requested injunctive relief and further ordered defendants to provide home tutoring for Doe on an interim basis; shortly thereafter, she issued a preliminary injunction directing defendants to return Doe to his then current educational placement at Louise Lombard School pending completion of the IEP review process. Doe reentered school on December 15, 514 weeks, and 24 schooldays, after his initial suspension.

Respondent Jack Smith was identified as an emotionally disturbed child by the time he entered the second grade in 1976. School records prepared that year indicated that he was unable “to control verbal or physical outburst[s]” and exhibited a “[sjevere disturbance in relationships with peers and adults.” Id., at 123. Further evaluations subsequently revealed that he had been physically and emotionally abused as an infant and young child and that, despite above average intelligence, he experienced academic and social difficulties as a result of extreme hyperactivity and low self-esteem. Id., at 136, 139, 155, 176. Of particular concern was Smith’s propensity for verbal hostility; one evaluator noted that the child reacted to stress by “attempting] to cover his feelings of low self worth through aggressive behavior[,]. . . primarily verbal provocations.” Id., at 136.

Based on these evaluations, SFUSD placed Smith in a learning center for emotionally disturbed children. His grandparents, however, believed that his needs would be better served in the public school setting and, in September 1979, the school district acceded to their requests and enrolled him at A. P. Giannini Middle School. His February 1980 IEP recommended placement in a Learning Disability Group, stressing the need for close supervision and a highly structured environment. Id., at 111. Like earlier evalúa*315tions, the February 1980 IEP noted that Smith was easily distracted, impulsive, and anxious; it therefore proposed a half-day schedule and suggested that the placement be undertaken on a trial basis. Id., at 112, 115.

At the beginning of the next school year, Smith was assigned to a full-day program; almost immediately thereafter he began misbehaving. School officials met twice with his grandparents in October 1980 to discuss returning him to a half-day program; although the grandparents agreed to the reduction, they apparently were never apprised of their right to challenge the decision through EHA procedures. The school officials also warned them that if the child continued his disruptive behavior — which included stealing, extorting money from fellow students, and making sexual comments to female classmates — they would seek to expel him. On November 14, they made good on this threat, suspending Smith for five days after he made further lewd comments. His principal referred the matter to the SPC, which recommended exclusion from SFUSD. As it did in John Doe’s case, the Committee scheduled a hearing and extended the suspension indefinitely pending a final disposition in the matter. On November 28, Smith’s counsel protested these actions on grounds essentially identical to those raised by Doe, and the SPC agreed to cancel the hearing and to return Smith to a half-day program at A. P. Giannini or to provide home tutoring. Smith’s grandparents chose the latter option and the school began home instruction on December 10; on January 6, 1981, an IEP team convened to discuss alternative placements.

After learning of Doe’s action, Smith sought and obtained leave to intervene in the suit. The District Court subsequently entered summary judgment in favor of respondents on their EHA claims and issued a permanent injunction. In a series of decisions, the District Judge found that the proposed expulsions and indefinite suspensions of respondents for conduct attributable to their disabilities deprived *316them of their congressionally mandated right to a free appropriate public education, as well as their right to have that education provided in accordance with the procedures set out in the EHA. The District Judge therefore permanently enjoined the school district from taking any disciplinary action other than a 2- or 5-day suspension against any disabled child for disability-related misconduct, or from effecting any other change in the educational placement of any such child without parental consent pending completion of any EHA proceedings. In addition, the judge barred the State from authorizing unilateral placement changes and directed it to establish an EHA compliance-monitoring system or, alternatively, to enact guidelines governing local school responses to disability-related misconduct. Finally, the judge ordered the State to provide services directly to disabled children when, in any individual case, the State determined that the local educational agency was unable or unwilling to do so.

On appeal, the Court of Appeals for the Ninth Circuit affirmed the orders with slight modifications. Doe v. Maher, 793 F. 2d 1470 (1986). Agreeing with the District Court that an indefinite suspension in aid of expulsion constitutes a prohibited “change in placement” under § 1415(e)(3), the Court of Appeals held that the stay-put provision admitted of no “dangerousness” exception and that the statute therefore rendered invalid those provisions of the California Education Code permitting the indefinite suspension or expulsion of disabled children for misconduct arising out of their disabilities. The court concluded, however, that fixed suspensions of up to 30 schooldays did not fall within the reach of § 1415(e)(3), and therefore upheld recent amendments to the state Education Code authorizing such suspensions.3 Lastly, the court *317affirmed that portion of the injunction requiring the State to provide services directly to a disabled child when the local educational agency fails to do so.

Petitioner Bill Honig, California Superintendent of Public Instruction,4 sought review in this Court, claiming that the Court of Appeals’ construction of the stay-put provision conflicted with that of several other Courts of Appeals which had recognized a dangerousness exception, compare Doe v. Maher, supra (case below), with Jackson v. Franklin County School Board, 765 F. 2d 535, 538 (CA5 1985); Victoria L. v. District School Bd. of Lee County, Fla., 741 F. 2d 369, 374 (CA11 1984); S-1 v. Turlington, 635 F. 2d 342, 348, n. 9 (CA5), cert. denied, 454 U. S. 1030 (1981), and that the direct services ruling placed an intolerable burden on the State. We granted certiorari to resolve these questions, 479 U. S. 1084 (1987), and now affirm.

II

At the outset, we address the suggestion, raised for the first time during oral argument, that this case is moot.5 Under Article III of the Constitution this Court may only adjudicate actual, ongoing controversies. Nebraska Press Assn. v. Stuart, 427 U. S. 539, 546 (1976); Preiser v. Newkirk, 422 U. S. 395, 401 (1975). That the dispute between the parties was very much alive when suit was filed, or at the time the Court of Appeals rendered its judgment, cannot substitute for the actual ease or controversy that an exercise of this Court’s jurisdiction requires. Steffel v. Thompson, *318415 U. S. 452, 459, n. 10 (1974); Roe v. Wade, 410 U. S. 113, 125 (1973). In the present case, we have jurisdiction if there is a reasonable likelihood that respondents will again suffer the deprivation of EHA-mandated rights that gave rise to this suit. We believe that, at least with respect to respondent Smith, such a possibility does in fact exist and that the case therefore remains justiciable.

Respondent John Doe is now 24 years old and, accordingly, is no longer entitled to the protections and benefits of the EHA, which limits eligibility to disabled children between the ages of 3 and 21. See 20 U. S. C. § 1412(2)(B). It is clear, therefore, that whatever rights to state educational services he may yet have as a ward of the State, see Tr. of Oral Arg. 23, 26, the Act would not govern the State’s provision of those services, and thus the case is moot as to him. Respondent Jack Smith, however, is currently 20 and has not yet completed high school. Although at present he is not faced with any proposed expulsion or suspension proceedings, and indeed no longer even resides within the SFUSD, he remains a resident of California and is entitled to a “free appropriate public education” within that State. His claims under the EHA, therefore, are not moot if the conduct he originally complained of is “ ‘capable of repetition, yet evading review.’” Murphy v. Hunt, 455 U. S. 478, 482 (1982). Given Smith’s continued eligibility for educational services under the EHA,6 the nature of his disability, and petitioner’s *319insistence that all local school districts retain residual authority to exclude disabled children for dangerous conduct, we have little difficulty concluding that there is a “reasonable *320expectation,” ibid., that Smith would once again be subjected to a unilateral “change in placement” for conduct growing out of his disabilities were it not for the statewide injunctive relief issued below.

Our cases reveal that, for purposes of assessing the likelihood that state authorities will reinflict a given injury, we generally have been unwilling to assume that the party seeking relief will repeat the type of misconduct that would once again place him or her at risk of that injury. See Los Angeles v. Lyons, 461 U. S. 95, 105-106 (1983) (no threat that party seeking injunction barring police use of chokeholds would be stopped again for traffic violation or other offense, or would resist arrest if stopped); Murphy v. Hunt, supra, at 484 (no reason to believe that party challenging denial of pretrial bail “will once again be in a position to demand bail”); O’Shea v. Littleton, 414 U. S. 488, 497 (1974) (unlikely that parties challenging discriminatory bond-setting, sentencing, and jury-fee practices would again violate valid criminal laws). No such reluctance, however, is warranted here. It is respondent Smith’s very inability to conform his conduct to socially acceptable norms that renders him “handicapped” within the meaning of the EHA. See 20 U. S. C. § 1401(1); 34 CFR § 300.5(b)(8) (1987). As noted above, the record is replete with evidence that Smith is unable to govern his aggressive, impulsive behavior — indeed, his notice of suspension acknowledged that “Jack’s actions seem beyond his control.” App. 152. In the absence of any suggestion that respondent has overcome his earlier difficulties, it is certainly reasonable to expect, based on his prior history of behavioral problems, that he will again engage in classroom misconduct. Nor is it reasonable to suppose that Smith’s future educational placement will so perfectly suit his emotional and academic needs that further disruptions on his part are improbable. Although Justice Sc alia suggests in his dissent, post, at 338, that school officials are unlikely to place Smith in a setting where they cannot control his misbehavior, any ef*321forts to ensure such total control must be tempered by the school system’s statutory obligations to provide respondent with a free appropriate public education in “the least restrictive environment,” 34 CFR § 300.552(d) (1987); to educate him, “to the maximum extent appropriate,” with children who are not disabled, 20 U. S. C. § 1412(5); and to consult with his parents or guardians, and presumably with respondent himself, before choosing a placement. §§ 1401(19), 1415 (b). Indeed, it is only by ignoring these mandates, as well as Congress’ unquestioned desire to wrest from school officials their former unilateral authority to determine the placement of emotionally disturbed children, see infra, at 323-324, that the dissent can so readily assume that respondent’s future placement will satisfactorily prevent any further dangerous conduct on his part. Overarching these statutory obligations, moreover, is the inescapable fact that the preparation of an IEP, like any other effort at predicting human behavior, is an inexact science at best. Given the unique circumstances and context of this case, therefore, we think it reasonable to expect that respondent will again engage in the type of misconduct that precipitated this suit.

We think it equally probable that, should he do so, respondent will again be subjected to the same unilateral school action for which he initially sought relief. In this regard, it matters not that Smith no longer resides within the SFUSD. While the actions of SFUSD officials first gave rise to this litigation, the District Judge expressly found that the lack of a state policy governing local school responses to disability-related misconduct had led to, and would continue to result in, EHA violations, and she therefore enjoined the state defendant from authorizing, among other things, unilateral placement changes. App. 247-248. She of course also issued injunctions directed at the local defendants, but they did not seek review of those orders in this Court. Only petitioner, the State Superintendent of Public Instruction, has invoked our jurisdiction, and he now urges us to hold that *322local school districts retain unilateral authority under the EHA to suspend or otherwise remove disabled children for dangerous conduct. Given these representations, we have every reason to believe that were it not for the injunction barring petitioner from authorizing such unilateral action, respondent would be faced with a real and substantial threat of such action in any California school district in which he enrolled. Cf. Los Angeles v. Lyons, supra, at 106 (respondent lacked standing to seek injunctive relief because he could not plausibly allege that police officers choked all persons whom they stopped, or that the city “authorized police officers to act in such manner” (emphasis added)). Certainly, if the SFUSD’s past practice of unilateral exclusions was at odds with state policy and the practice of local school districts generally, petitioner would not now stand before us seeking to defend the right of all local school districts to engage in such aberrant behavior.7

We have previously noted that administrative and judicial review under the EHA is often “ponderous,” Burlington School Committee v. Massachusetts Dept. of Education, 471 U. S. 359, 370 (1985), and this case, which has taken seven years to reach us, amply confirms that observation. For obvious reasons, the misconduct of an emotionally disturbed or otherwise disabled child who has not yet reached adolescence typically will not pose such a serious threat to the well-being of other students that school officials can only ensure classroom safety by excluding the child. Yet, the adolescent student improperly disciplined for misconduct that does pose such a threat will often be finished with school or otherwise *323ineligible for EHA protections by the time review can be had in this Court. Because we believe that respondent Smith has demonstrated both “a sufficient likelihood that he will again be wronged in a similar way,” Los Angeles v. Lyons, 461 U. S., at 111, and that any resulting claim he may have for relief will surely evade our review, we turn to the merits of his case.

Ill

The language of § 1415(e)(3) is unequivocal. It states plainly that during the pendency of any proceedings initiated under the Act, unless the state or local educational agency and the parents or guardian of a disabled child otherwise agree, “the child shall remain in the then current educational placement.” § 1415(e)(3) (emphasis added). Faced with this clear directive, petitioner asks us to read a “dangerousness” exception into the stay-put provision on the basis of either of two essentially inconsistent assumptions: first, that Congress thought the residual authority of school officials to exclude dangerous students from the classroom too obvious for comment; or second, that Congress inadvertently failed to provide such authority and this Court must therefore remedy the oversight. Because we cannot accept either premise, we decline petitioner’s invitation to rewrite the statute.

Petitioner’s arguments proceed, he suggests, from a simple, commonsense proposition: Congress could not have intended the stay-put provision to be read literally, for such a construction leads to the clearly unintended, and untenable, result that school districts must return violent or dangerous students to school while the often lengthy EHA proceedings run their course. We think it clear, however, that Congress very much meant to strip schools of the unilateral authority they had traditionally employed to exclude disabled students, particularly emotionally disturbed students, from school. In so doing, Congress did not leave school administrators powerless to deal with dangerous students; it did, however, deny school officials their former right to “self-help,” and directed *324that in the future the removal of disabled students could be accomplished only with the permission of the parents or, as a last resort, the courts.

As noted above, Congress passed the EHA after finding that school systems across the country had excluded one out of every eight disabled children from classes. In drafting the law, Congress was largely guided by the recent decisions in Mills v. Board of Education of District of Columbia, 348 F. Supp. 866 (1972), and PARC, 343 F. Supp. 279 (1972), both of which involved the exclusion of hard-to-handle disabled students. Mills in particular demonstrated the extent to which schools used disciplinary measures to bar children from the classroom. There, school officials had labeled four of the seven minor plaintiffs “behavioral problems,” and had excluded them from classes without providing any alternative education to them or any notice to their parents. 348 F. Supp., at 869-870. After finding that this practice was not limited to the named plaintiffs but affected in one way or another an estimated class of 12,000 to 18,000 disabled students, id., at 868-869, 875, the District Court enjoined future exclusions, suspensions, or expulsions “on grounds of discipline.” Id., at 880.

Congress attacked such exclusionary practices in a variety of ways. It required participating States to educate all disabled children, regardless of the severity of their disabilities, 20 U. S. C. § 1412(2)(C), and included within the definition of “handicapped” those children with serious emotional disturbances. § 1401(1). It further provided for meaningful parental participation in all aspects of a child’s educational placement, and barred schools, through the stay-put provision, from changing that placement over the parent’s objection until all review proceedings were completed. Recognizing that those proceedings might prove long and tedious, the Act’s drafters did not intend § 1415(e)(3) to operate inflexibly, see 121 Cong. Rec. 37412 (1975) (remarks of Sen. Stafford), and they therefore allowed for interim placements where par*325ents and school officials are able to agree on one. Conspicuously absent from § 1415(e)(3), however, is any emergency exception for dangerous students. This absence is all the more telling in light of the injunctive decree issued in PARC, which permitted school officials unilaterally to remove students in “‘extraordinary circumstances.’” 343 F. Supp., at 301. Given the lack of any similar exception in Mills, and the close attention Congress devoted to these “landmark” decisions, see S. Rep., at 6, we can only conclude that the omission was intentional; we are therefore not at liberty to en-graft onto the statute an exception Congress chose not to create.

Our conclusion that § 1415(e)(3) means what it says does not leave educators hamstrung. The Department of Education has observed that, “[w]hile the [child’s] placement may not be changed [during any complaint proceeding], this does not preclude the agency from using its normal procedures for dealing with children who are endangering themselves or others.” Comment following 34 CFR §300.513 (1987). Such procedures may include the use of study carrels, timeouts, detention, or the restriction of privileges. More drastically, where a student poses an immediate threat to the safety of others, officials may temporarily suspend him or her for up to 10 schooldays.8 This authority, which respondent *326in no way disputes, not only ensures that school administrators can protect the safety of others by promptly removing the most dangerous of students, it also provides a “cooling down” period during which officials can initiate IEP review and seek to persuade the child’s parents to agree to an interim placement. And in those cases in which the parents of a truly dangerous child adamantly refuse to permit any change in placement, the 10-day respite gives school officials an opportunity to invoke the aid of the courts under § 1415(e)(2), which empowers courts to grant any appropriate relief.

Petitioner contends, however, that the availability of judicial relief is more illusory than real, because a party seeking review under § 1415(e)(2) must exhaust time-consuming administrative remedies, and because under the Court of Appeals’ construction of § 1415(e)(3), courts are as bound by the stay-put provision’s “automatic injunction,” 793 F. 2d, at 1486, as are schools.9 It is true that judicial review is nor*327mally not available under § 1415(e)(2) until all administrative proceedings are completed, but as we have previously noted, parents may bypass the administrative process where exhaustion would be futile or inadequate. See Smith v. Robinson, 468 U. S. 992, 1014, n. 17 (1984) (citing cases); see also 121 Cong. Rec. 37416 (1975) (remarks of Sen. Williams) (“[E]xhaustion. . . should not be required ... in cases where such exhaustion would be futile either as a legal or practical matter”). While many of the EHA’s procedural safeguards protect the rights of parents and children, schools can and do seek redress through the administrative review process, and we have no reason to believe that Congress meant to require schools alone to exhaust in all cases, no matter how exigent the circumstances. The burden in such cases, of course, rests with the school to demonstrate the futility or inadequacy of administrative review, but nothing in § 1415(e)(2) suggests that schools are completely barred from attempting to make such a showing. Nor do we think that § 1415(e)(3) operates to limit the equitable powers of district courts such that they cannot, in appropriate cases, temporarily enjoin a dangerous disabled child from attending school. As the EHA’s legislative history makes clear, one of the evils Congress sought to remedy was the unilateral exclusion of disabled children by schools, not courts, and one of the purposes of § 1415(e)(3), therefore, was “to prevent school officials from removing a child from the regular public school classroom over the parents’ objection pending completion of the review proceedings.” Burlington School Committee v. Massachusetts Dept, of Education, 471 U. S., at 373 (emphasis added). The stay-put provision in no way purports to limit or pre-empt the authority conferred on courts by § 1415(e)(2), see Doe v. Brookline School Committee, 722 F. 2d 910, 917 (CA1 1983); indeed, it says nothing whatever about judicial power.

*328In short, then, we believe that school officials are entitled to seek injunctive relief under § 1415(e)(2) in appropriate cases. In any such action, § 1415(e)(3) effectively creates a presumption in favor of the child’s current educational placement which school officials can overcome only by showing that maintaining the child in his or her current placement is substantially likely to result in injury either to himself or herself, or to others. In the present case, we are satisfied that the District Court, in enjoining the state and local defendants from indefinitely suspending respondent or otherwise unilaterally altering his then current placement, properly balanced respondent’s interest in receiving a free appropriate public education in accordance with the procedures and requirements of the EH A against the interests of the state and local school officials in maintaining a safe learning environment for all their students.10

IV

We believe the courts below properly construed and applied § 1415(e)(3), except insofar as the Court of Appeals held that a suspension in excess of 10 schooldays does not consti*329tute a “change in placement.”11 We therefore affirm the Court of Appeals’ judgment on this issue as modified herein. Because we are equally divided on the question whether a court may order a State to provide services directly to a disabled child where the local agency has failed to do so, we affirm the Court of Appeals’ judgment on this issue as well.

Affirmed.

concurring.

I write separately on the mootness issue in this case to explain why I have joined Part II of the Court’s opinion, and why I think reconsideration of our mootness jurisprudence may be in order when dealing with cases decided by this Court.

The present rule in federal cases is that an actual controversy must exist at all stages of appellate review, not merely at the time the complaint is filed. This doctrine was clearly articulated in United States v. Munsingwear, Inc., 340 U. S. 36 (1950), in which Justice Douglas noted that “[t]he established practice of the Court in dealing with a civil case from a court in the federal system which has become moot while on its way here or pending our decision on the merits is to reverse or vacate the judgment below and remand with a direction to dismiss.” Id., at 39. The rule has been followed fairly consistently over the last 30 years. See, e. g., Preiser v. Newkirk, 422 U. S. 395 (1975); SEC v. Medical Committee for Human Rights, 404 U. S. 403 (1972).

All agree that this case was “very much alive,” ante, at 317, when the action was filed in the District Court, and very probably when the Court of Appeals decided the case. It is supervening events since the decision of the Court of Appeals which have caused the dispute between the majority and the dissent over whether this case is moot. Therefore, all that the Court actually holds is that these supervening events do *330not deprive this Court of the authority to hear the case. I agree with that holding, and would go still further in the direction of relaxing the test of mootness where the events giving rise to the claim of mootness have occurred after our decision to grant certiorari or to note probable jurisdiction.

The Court implies in its opinion, and the dissent expressly states, that the mootness doctrine is based upon Art. Ill of the Constitution. There is no doubt that our recent cases have taken that position. See Nebraska Press Assn. v. Stuart, 427 U. S. 539, 546 (1976); Preiser v. Newkirk, supra, at 401; Sibron v. New York, 392 U. S. 40, 57 (1968); Liner v. Jafco, Inc., 375 U. S. 301, 306, n. 3 (1964). But it seems very doubtful that the earliest case I have found discussing mootness, Mills v. Green, 159 U. S. 651 (1895), was premised on constitutional constraints; Justice Gray’s opinion in that case nowhere mentions Art. III.

If it were indeed Art. Ill which — by reason of its requirement of a case or controversy for the exercise of federal judicial power — underlies the mootness doctrine, the “capable of repetition, yet evading review” exception relied upon by the Court in this case would be incomprehensible. Article III extends the judicial power of the United States only to cases and controversies; it does not except from this requirement other lawsuits which are “capable of repetition, yet evading review.” If our mootness doctrine were forced upon us by the case or controversy requirement of Art. Ill itself, we would have no more power to decide lawsuits which are “moot” but which also raise questions which are capable of repetition but evading review than we would to decide cases which are “moot” but raise no such questions.

The exception to mootness for cases which are “capable of repetition, yet evading review,” was first stated by this Court in Southern Pacific Terminal Co. v. ICC, 219 U. S. 498 (1911). There the Court enunciated the exception in the light of obvious pragmatic considerations, with no mention of Art. Ill as the principle underlying the mootness doctrine:

*331“The questions involved in the orders of the Interstate Commerce Commission are usually continuing (as are manifestly those in the case at bar) and their consideration ought not to be, as they might be, defeated, by short term orders, capable of repetition, yet evading review, and at one time the Government and at another time the carriers have their rights determined by the Commission without a chance of redress.” Id., at 515.

The exception was explained again in Moore v. Ogilvie, 394 U. S. 814, 816 (1969):

“The problem is therefore ‘capable of repetition, yet evading review.’ The need for its resolution thus reflects a continuing controversy in the federal-state area where our ‘one man, one vote’ decisions have thrust” (citation omitted).

It is also worth noting that Moore v. Ogilvie involved a question which had been mooted by an election, just as did Mills v. Green some 74 years earlier. But at the time of Mills, the case originally enunciating the mootness doctrine, there was no thought of any exception for cases which were “capable of repetition, yet evading review.”

The logical conclusion to be drawn from these cases, and from the historical development of the principle of mootness, is that while an unwillingness to decide moot cases may be connected to the case or controversy requirement of Art. Ill, it is an attenuated connection that may be overridden where there are strong reasons to override it. The “capable of repetition, yet evading review” exception is an example. So too is our refusal to dismiss as moot those cases in which the defendant voluntarily ceases, at some advanced stage of the appellate proceedings, whatever activity prompted the plaintiff to seek an injunction. See, e. g., City of Mesquite v. Aladdin’s Castle, Inc., 455 U. S. 283, 289, n. 10 (1982); United States v. W. T. Grant Co., 345 U. S. 629, 632 (1953). I believe that we should adopt an additional exception to our *332present mootness doctrine for those cases where the events which render the case moot have supervened since our grant of certiorari or noting of probable jurisdiction in the case. Dissents from denial of certiorari in this Court illustrate the proposition that the roughly 150 or 160 cases which we decide each year on the merits are less than the number of cases warranting review by us if we are to remain, as Chief Justice Taft said many years ago, “the last word on every important issue under the Constitution and the statutes of the United States.” But these unique resources — the time spent preparing to decide the case by reading briefs, hearing oral argument, and conferring — are squandered in every case in which it becomes apparent after the decisional process is underway that we may not reach the question presented. To me the unique and valuable ability of this Court to decide a case — we are, at present, the only Art. Ill court which can decide a federal question in such a way as to bind all other courts —is a sufficient reason either to abandon the doctrine of mootness altogether in cases which this Court has decided to review, or at least to relax the doctrine of mootness in such a manner as the dissent accuses the majority of doing here. I would leave the mootness doctrine as established by our cases in full force and effect when applied to the earlier stages of a lawsuit, but I believe that once this Court has undertaken a consideration of a case, an exception to that principle is just as much warranted as where a case is “capable of repetition, yet evading review.”