Restraint & seclusion in U.S. schools: what the federal data shows.
The U.S. Department of Education records how public schools physically restrain, mechanically restrain, and seclude students through its Civil Rights Data Collection (CRDC). Prior Signal reproduced two collection cycles of that record — 2017–18 and 2021–22 — directly from the Department’s published files and rebuilt them so anyone can read them.
This is an independent analysis of public federal records. Figures reproduce the U.S. Department of Education’s Civil Rights Data Collection and are cited in full below; characterizations are Prior Signal’s.
Two findings from the federal record stand out.
Prior Signal analyzed federal Civil Rights Data Collection records covering physical restraint, mechanical restraint, and seclusion across 98,010 public schools in the 2021–22 cycle.
of students physically restrained were students with disabilities (IDEA + Section 504) — a group that is about 14–16% of enrollment. [2]
increase in mechanical restraint of students, from 3,619 in 2017–18 to 8,219 in 2021–22. [2][3][3a]
Sources: U.S. Department of Education, Office for Civil Rights, Civil Rights Data Collection (CRDC), 2021–22 and 2017–18 school years; “2021–22 CRDC: A First Look”, released January 2025. Full citations below. [1][2][3]
How many students are restrained or secluded in U.S. schools?
The federal record is the Civil Rights Data Collection, gathered by the U.S. Department of Education from nearly every public school in the country. It tracks three distinct practices:
Physical restraint
A personal restriction that immobilizes or reduces the ability of a student to move his or her torso, arms, legs, or head freely. It does not include a physical escort — a temporary touching or holding to guide a student to a safe location. [6]
Mechanical restraint
The use of any device or equipment to restrict a student’s freedom of movement, including handcuffs or similar devices used by law enforcement or school security. It does not include devices prescribed by a medical or related-services professional for approved purposes, such as adaptive devices, vehicle safety restraints, or orthopedically prescribed devices. [6]
Seclusion
The involuntary confinement of a student in a room or area, with or without adult supervision, from which the student is not permitted to leave. It does not include a classroom environment requiring routine permission to leave, an approved behavior-management technique in an unlocked setting, or placing a student in a separate location within a classroom with the same access to instruction. [6]
The data has been public for years. The problem was never access; it was legibility. Raw federal files are not something a parent, teacher, advocate, or reporter can open and understand. Prior Signal reproduced the record from the primary source, then made it readable.
Disability overrepresentation is the through-line.
The single most consistent pattern in the data is disability overrepresentation. In the 2021–22 CRDC, students with disabilities (IDEA + Section 504) accounted for 77.3% of students subjected to physical restraint, despite representing about 14–16% of total enrollment. [2]
A group that is about one in seven students by headcount is more than three in four students by physical restraint. That gap is the story. It holds across the record, and it is the finding advocates return to most.
The U.S. Department of Education’s own headline figure, using IDEA alone (not IDEA + Section 504), is close: in its 2021–22 “First Look” report, students served under IDEA represented 14% of K–12 enrollment but 28% of students mechanically restrained, 68% of students secluded, and 76% of students physically restrained. Prior Signal’s 77.3% figure reproduces the same official school-level file with the broader IDEA + Section 504 disability definition; both are correct, sourced, and shown here. [2]
Mechanical restraint more than doubled since 2017–18.
Mechanical restraint is the smallest of the three categories the CRDC tracks, but it moved the most. Nationally, the number of students mechanically restrained rose from 3,619 in 2017–18 to 8,219 in 2021–22 — an increase of about 127%. [2][3][3a] Physical restraint (70,833 → 68,737) and seclusion (27,538 → 28,712) were roughly flat over the same span, which is why mechanical restraint stands out. Two caveats still apply: mechanical restraint is the smallest-base measure, so percentage swings are naturally larger; and the U.S. Government Accountability Office has found the Department of Education’s quality-control processes for this data “largely ineffective or do not exist,” which affects confidence in exact year-over-year comparisons even where the reported counts themselves are accurately reproduced here. [4] The 2020–21 pandemic-year collection (~2,100 mechanically restrained) is excluded from this comparison; see the Method note.
Every figure reproduces the official record.
Every figure on this page and the linked visualization reproduces the U.S. Department of Education’s official Civil Rights Data Collection. Prior Signal does not model, smooth, or estimate these values; it reproduces the government’s published figures and makes them legible. Where a value was suppressed in the federal public-use file for student-privacy reasons, it is omitted here rather than imputed.
Prior Signal reproduced two CRDC cycles: 2017–18 and 2021–22. The 2020–21 collection is deliberately excluded from trend comparisons because it was gathered during the COVID-19 pandemic, when most students were learning remotely and reported restraint and seclusion fell to roughly 52,800 students nationally — an anomaly that would distort any trend. [5]
98,010 is the number of public schools covered by the 2021–22 CRDC as reported by the U.S. Department of Education. [2]
Percentages are student-count measures from the federal record, not incident-count estimates.
The page reports only values that could be traced back to the government’s own files and verified before publication.
These practices are known to be under-counted at the source. The U.S. Government Accountability Office found the Department of Education’s quality-control processes for restraint and seclusion data “largely ineffective or do not exist,” and that in the 2015–16 collection about 70% of school districts reported zero incidents — including large districts later shown to have had incidents. The figures reproduced here therefore represent a floor, not a ceiling: the true numbers are very likely higher. [4][5a]
Who built this, and why?
Prior Signal built this visualization using two years of Civil Rights Data Collection data, in collaboration with the Alliance Against Seclusion & Restraint — an organization that works to end the physical restraint and seclusion of children in schools. The organization had the federal data but no practical way for the public to read it. Prior Signal closed that gap.
The method did not start in the United States. It started with End The Wait Ontario, a public record built from freedom-of-information responses and government data on Ontario’s autism waitlist, with every figure traceable to a primary source. That project proved the discipline: evidence earns trust when it is readable, source-grounded, and easy to check.
Different country. Same mission: evidence that holds up, in the hands of the people using it to change something.
Frequently asked questions.
What is the Civil Rights Data Collection?
The CRDC is a survey conducted by the U.S. Department of Education that gathers data from public schools on civil-rights indicators, including the physical restraint, mechanical restraint, and seclusion of students. It is the authoritative federal source on these practices.
What is the difference between restraint and seclusion?
Restraint restricts a student’s movement, either physically by a person or mechanically by a device. Seclusion confines a student alone in a space they are prevented from leaving. They are distinct practices tracked separately in the federal data.
Are students with disabilities restrained more often?
Yes. In the 2021–22 CRDC, students with disabilities served under IDEA made up about 14% of enrollment but 76% of students physically restrained — a substantial overrepresentation that the U.S. Department of Education reports directly.
Is this restraint and seclusion data reliable?
Every figure reproduces the U.S. Department of Education’s Civil Rights Data Collection exactly as published; nothing here is estimated or modeled. One important caveat applies to the underlying federal data itself: the U.S. Government Accountability Office has found that these practices are under-reported, with quality-control processes it described as “largely ineffective or do not exist.” Prior Signal reproduces the official record faithfully; because that record undercounts, the real-world totals are very likely higher than the figures shown.
Who is Prior Signal?
Prior Signal is an OSINT and strategic-intelligence firm that turns public records into verified, legible, citable evidence for organizations that cannot afford to be wrong.
Sources & citations.
Data currency: 2021–22 CRDC (released January 2025) is the latest national cycle. Last verified: 2026-07-08.
- [1] U.S. Department of Education, Office for Civil Rights, Civil Rights Data Collection (CRDC). https://civilrightsdata.ed.gov/
- [2] 2021–22 Civil Rights Data Collection — A First Look: Students’ Access to Educational Opportunities, released January 2025. Reports: 98,010 public schools; students with disabilities served under IDEA were 14% of K-12 enrollment but 76% of students physically restrained, 28% of students mechanically restrained, and 68% of students secluded; approximately 68,800 students physically restrained, 8,200 mechanically restrained, and 28,700 secluded. https://www.ed.gov/media/document/2021-22-crdc-first-look-report-109194.pdf
- [3] Civil Rights Data Collection, 2017–18 school year — Restraint and Seclusion Issue Brief (issued October 15, 2020) and the 2017–18 CRDC State and National Estimations table. Reports 101,990 students subjected to restraint or seclusion in 2017–18 nationally (physical restraint 70,833; mechanical restraint 3,619; seclusion 27,538); of students restrained, 78% were students with disabilities; of students secluded, 77% were students with disabilities. https://www.ed.gov/laws-and-policy/civil-rights-laws/civil-rights-data-collection-crdc/civil-rights-data/civil-rights-data-collection-crdc-2017-18-school-year
- [3a] 2017–18 CRDC State and National Estimations table (interactive), the source for the 2017–18 mechanical-restraint national total (3,619) used in the trend comparison above. https://civilrightsdata.ed.gov/estimations/2017-2018
- [4] U.S. Government Accountability Office, K-12 Education: Education Needs to Address Significant Quality Issues with its Restraint and Seclusion Data, GAO-20-345 (April 2020). Finds the Department of Education’s CRDC quality-control processes for restraint and seclusion data "largely ineffective or do not exist." https://www.gao.gov/products/gao-20-345
- [5a] U.S. Government Accountability Office, K-12 Education: Education Should Take Immediate Action to Address Inaccuracies in Federal Restraint and Seclusion Data, GAO-19-551R (June 18, 2019; reissued July 11, 2019). Finds that in the 2015–16 CRDC about 70% of districts reported zero incidents, and nine of the 30 largest districts inaccurately reported zeros. https://www.gao.gov/products/gao-19-551r
- [5] 2020–21 CRDC — A First Look (released November 2023). Source for the 2020–21 pandemic-year anomaly figure (~52,800 students restrained or secluded). https://www.ed.gov/media/document/crdc-educational-opportunities-reportpdf-21412.pdf
- [6] Definitions: U.S. Department of Education CRDC Restraint & Seclusion module, key definitions (physical restraint, mechanical restraint, seclusion). https://crdc.communities.ed.gov/sites/default/files/2024-12/RSTR%20Individual%20Module-Restraint%20&%20Seclusion.pdf
- [7] 17 U.S.C. § 105 — Copyright status of U.S. Government works. https://www.law.cornell.edu/uscode/text/17/105
Source data: U.S. Department of Education, Civil Rights Data Collection — a work of the U.S. federal government, not subject to copyright (17 U.S.C. § 105). Analysis, compilation, and visualization © 2026 Prior Signal.
If you are holding a public record no one can read, make it evidence that holds up.
Prior Signal turns data collections, freedom-of-information releases, registries, and document troves into evidence people can read, check, and cite.