Subjects and lessons to explore.

Law

Criminal Justice & Criminology · Corrections

Rehabilitation and Reentry

13 min read 5 practice questions
10 key termsSourcesReport an error
Want it in plain words first? Jump to Eli explains — the same idea, no jargon.
On this page 9 sections
  1. In 30 seconds
  2. Why this matters
  3. The college version
  4. Eli explains
  5. Worked example
  6. Key takeaway
  7. Quick check
  8. Study tools
  9. Sources & references

In 30 seconds

is the claim that correctional programs can lower future offending. is the practical problem behind the 453,242 releases of sentenced prisoners U.S. systems recorded in 2023. A 1974 review was compressed into the slogan nothing works; its author withdrew that conclusion in 1979. The evidence since says well-run programs matched to risk produce modest reductions, and that rearrest, reconviction, and return to prison are three different numbers that studies report under one word.

Why this matters

Almost everyone sent to prison comes back out, so what happens during and after a sentence is a design question, not a philosophical one. This is also the corner of criminal justice where numbers are most often misused: a single word, , hides at least three measures and follow-up windows that range from one year to ten. Learning to ask which measure, over how long, for which released population, is the skill that separates a defensible claim from a talking point. Courses in corrections, program evaluation, and policy analysis all build on it, and the same habit transfers to any field where a program claims to have reduced a bad outcome. This lesson is educational material about how the U.S. system and its research literature work. It is not legal, benefits, or medical advice, and it takes no position on what policy should be.

The college version

The rehabilitative ideal and the survey that unsettled it

For most of the twentieth century, American corrections was organized around the premise that confinement could change people. Indeterminate sentences, parole boards, prison schools, and treatment programs all assumed a person could be assessed, treated, and released when improved. In the late 1960s a New York State governor's committee, working from exactly that premise, hired a research team to find out which methods actually worked. The team read the evaluation literature published between the end of World War II and 1967 and kept the 231 studies that met their design standards: an evaluation of a treatment method, an independent measure of improvement, and some untreated comparison group. The resulting manuscript ran roughly 1,400 pages, and the state declined to publish it. Robert Martinson, one of the three authors, published a short account of the recidivism findings in the Spring 1974 issue of The Public Interest under the title "What works?" The article, not the manuscript, is what the field remembers, and it is remembered mostly wrong.

What Martinson wrote, and what he wrote five years later

Martinson's summary sentence was that with few and isolated exceptions, the rehabilitative efforts reported so far had had "no appreciable effect on recidivism." He called that an unqualified statement and then spent the article fielding objections to it. Three qualifications matter. First, he warned before the sentence that studies using the words recidivism rate were often measuring different things, arrest rates or parole violation rates or favorable discharge, and that those measures do not correlate highly. Second, he wrote near the end that the research base was riddled with unreplicated experiments and follow-up periods that rarely extended past legal supervision, and that it was possible some programs were working but the research was too weak to detect it. Third, he did not claim nothing reduces crime; he speculated that deterrence might be doing work no one had measured. In 1979, in the Hofstra Law Review, he withdrew the conclusion outright, withdrew his earlier description of treatment as impotent, and reported from a larger survey that programs could be beneficial, neutral, or harmful depending on the conditions under which they were delivered. The retraction never caught up with the slogan.

What works, restated: risk, need, responsivity

The productive response to "nothing works" was not louder advocacy but better synthesis. The framework that emerged, formalized in 1990 by Andrews, Bonta, and Hoge and summarized in a 2007 Public Safety Canada report by Bonta and Andrews, is risk-need-responsivity. The says match the intensity of service to the person's assessed risk of reoffending: reserve intensive programming for higher-risk cases. The need principle says target criminogenic needs, the changeable risk factors actually linked to offending, rather than needs that matter for wellbeing but not for reoffending. Andrews and Bonta group the major predictors into the : antisocial personality pattern, procriminal attitudes, social supports for crime, substance abuse, family and marital relationships, school and work, prosocial recreation, and criminal history, the last of which is static and cannot be treated. The says deliver the intervention using cognitive social learning methods and tune it to the individual's motivation, abilities, and learning style. Together the three answer who, what, and how.

The reported effect sizes, and their heterogeneity

Bonta and Andrews report the accumulated meta-analytic results as differences in recidivism between treated and comparison groups. Across 374 tests of the risk principle, treatment delivered to higher-risk cases was associated with about a 10 percent difference in recidivism, while treatment delivered to lower-risk cases produced about 3 percent. Targeting criminogenic needs was associated with about a 19 percent difference; targeting non-criminogenic needs was associated with roughly a 1 percent increase. Adherence to general responsivity alone was associated with about a 23 percent difference. Programs adhering to all three principles showed average differences of about 17 percent in residential or custodial settings and about 35 percent in community settings, while programs adhering to none were associated with increased recidivism, an effect the authors report as worse in custody. These are averages across heterogeneous studies with varying designs, outcome measures, and follow-up windows, and they describe program-level differences, not guarantees for an individual. Read as a distribution rather than a promise, they support a modest claim: some programs help, some do nothing, and poorly matched programs can make things worse.

Implementation is part of the treatment

The most durable finding in this literature is not about program type. Bonta and Andrews report that treatment delivered in ordinary agency conditions achieves roughly half the effect of the same treatment delivered in a controlled demonstration, and that the accuracy of staff risk assessments drifts over time without maintenance. Two facilities can run the same branded curriculum and get different results because one screens participants by assessed risk, trains and supervises facilitators, and completes the dosage while the other does not. This is why evidence syntheses in corrections talk about , fidelity, and dosage alongside content, and why a rating attached to a practice cannot be transferred to a local program that shares only its name. When you read that a program "works," ask what was implemented, for whom, at what intensity, and whether anyone verified it.

The scale of release

Reentry is not a niche. States and the Federal Bureau of Prisons reported 453,242 releases of sentenced prisoners in 2023, of which 408,521 were from state systems and 44,721 from the federal system, against 472,300 admissions the same year. About 70 percent of those releases were conditional on community supervision or other requirements, which is why parole and probation supervision is the default endpoint of a prison sentence rather than the exception. Two limits on that number matter. It counts prisoners serving sentences of more than one year, so it is not a count of everyone who leaves confinement. And it excludes local jails entirely, where turnover is far higher and stays are short: BJS reports that people admitted to local jails between July 2022 and June 2023 spent an average of 32 days in custody before release. Date any release or population figure you cite and say which population it covers.

Recidivism is not one number

This is the methodological point that governs everything else in the topic. NIJ defines recidivism as a return to criminal behavior after a sanction or intervention, and notes it is typically measured by rearrest, reconviction, or reincarceration over a stated . Those are different events with different base rates, and studies pick different ones. The three landmark BJS studies show the range. Among state prisoners released in 30 states in 2005 and followed nine years, 68 percent were arrested within 3 years, 79 percent within 6, and 83 percent within 9. Among prisoners released in 34 states in 2012 and followed five years, 62 percent were arrested within 3 years and 71 percent within 5, while 46 percent returned to prison within 5. Among prisoners released in 24 states in 2008 and followed ten years, 66 percent were arrested within 3 years and 82 percent within 10, while 61 percent returned to prison within 10 for a parole or probation violation or a new sentence. Note what that last measure blends: a return to prison can follow a new crime or a supervision violation, so it is partly a measure of how a state supervises rather than of new offending alone. Follow-up length matters as much as measure choice; BJS found the annual arrest percentage in the 2008 cohort fell from 43 percent in year one to 22 percent in year ten. Two honest studies of the same population can report 61 percent and 82 percent.

The first weeks: documents, housing, work, coverage, debt

The practical problem of release is that several systems need proof of identity and stable address at once, and a person leaving custody may have neither. Some states legislate around the first problem: California requires its corrections department and the Department of Motor Vehicles to ensure that eligible people released from state prison obtain a valid state identification card, and directs staff to help obtain the underlying documents. Housing assistance is conditioned in federal regulation; the Housing Choice Voucher rules require housing agencies to deny admission to a household containing a member subject to lifetime sex-offender registration or convicted of manufacturing methamphetamine in federally assisted housing, and permit denial on other criminal-history grounds. On employment, EEOC enforcement guidance treats criminal-record screens as subject to Title VII disparate-impact analysis, and treats an exclusion based on an arrest alone as not job related and consistent with business necessity. Health coverage breaks at both ends of a sentence because federal Medicaid law excludes payment for services to an inmate of a public institution except as a patient in a medical institution, so re-establishing coverage is its own task at release. Court debt, fees, and restitution typically survive the sentence. How the classification of the conviction generates these consequences as a matter of law belongs to the offense-classification topic; what matters here is that they land at once, in the weeks when a person has the least capacity to absorb them, and that state law varies.

Reading the program evidence honestly

NIJ's CrimeSolutions clearinghouse rates practices and programs against the underlying evaluations, which makes it a better starting point than advocacy summaries because it publishes nulls. Its Adult Reentry Programs profile is rated Promising: a meta-analysis of 53 studies found a statistically significant effect on recidivism of r = .06, which the profile translates as roughly 47 percent recidivism for participants against 53 percent for comparisons, while a second meta-analysis limited to randomized trials of adult male samples found no statistically significant effect on rearrest. Its Rehabilitation Programs for Adults profile is also rated Promising, based on Lipsey's 2019 meta-analysis of 801 studies, and reports that structured group work, cognitive behavioral components, counseling, and specialized courts were associated with statistically significant recidivism reductions while work-related, academic, supportive residential, intensive supervision, multimodal, and restorative interventions were not. Individual programs can fail outright: an Oklahoma juvenile reentry program funded under the Second Chance Act is rated Ineffective after an evaluation found no statistically significant difference in reconviction at 18 months. Both practice profiles carry a notice that they are no longer being updated, which is itself a reason to check the underlying studies.

Correctional education and the federal funding vehicle

Correctional education has the comparatively strongest evidence base in this area, and it is worth stating precisely. RAND's 2013 meta-analysis for the Bureau of Justice Assistance found that people who participated in correctional education programs had 43 percent lower odds of recidivating than those who did not, which the authors translate into a 13 percentage-point reduction in risk, with the result stable when lower-quality studies were included. High school or GED programs alone showed 30 percent lower odds. Employment results were weaker: 13 percent higher odds of postrelease employment overall, but only one study met the higher-quality threshold, so the authors decline to claim a statistically significant employment effect. They also report that they could not disentangle program types and found methodological weaknesses across the literature that limit policy inference. Studies rating 2 or higher on the Maryland Scientific Methods Scale were eligible, so non-randomized comparisons of participants to nonparticipants qualified, and selection into programs remains a live threat. On funding, the Second Chance Act of 2007, Public Law 110-199, signed April 9, 2008, is the federal grant vehicle: it authorizes adult and juvenile reentry demonstration projects, reentry courts, mentoring grants, and substance-abuse programming, and it was amended by the First Step Act of 2018. Its purposes section is codified at 34 U.S.C. 60501. A funding stream is not an outcome; the Oklahoma program above was funded under it and did not work.

Eli, the EliExplains learning guide

Eli explains

The same idea, in plain words

Explain it like I’m 10

Imagine a school claims its tutoring program helps students. To check, you have to ask three things. Which students got the tutoring, since tutoring a student who was already fine will not show much. What counts as helping, since "missed fewer classes" and "passed the year" are different tests and give different numbers. And how long you watched, since checking after one month and checking after five years give very different answers about the same students. Corrections research has exactly these three problems. In 1974 a researcher summarized a pile of prison program studies and said, roughly, none of this shows a clear effect. Newspapers turned that into "nothing works." He himself said five years later that the conclusion was wrong and that programs could help, do nothing, or make things worse depending on how they were run. Today the honest version is: careful programs, aimed at the people most likely to reoffend, working on the habits and situations actually tied to offending, delivered properly, produce real but modest improvements. And leaving prison is its own separate problem, because a person may need an ID card, an address, a job, and health coverage in the same week.

Picture it like this

Think of a hospital discharge. Treatment inside the building is only half the job; if the patient leaves with no prescription, no follow-up appointment, and no way to get to the pharmacy, the treatment can be excellent and the outcome can still be bad.

Where the picture stops working

The analogy breaks down in three places. A hospital discharge is voluntary and a prison release usually is not, and about 70 percent of 2023 prison releases carried supervision conditions the person did not choose. A patient is not legally barred from housing or jobs because of a diagnosis, while a conviction triggers rules that restrict both. And offending is not an illness with a known cause and cure; the risk factors researchers target are statistical predictors, not diagnoses, and the strongest programs shift group averages rather than curing individuals.

Worked example

Take the BJS cohort of prisoners released in 24 states in 2008 and followed for ten years. Someone writes that this cohort had a recidivism rate of 82 percent. Someone else writes 61 percent. Both are reading the same report. The first is using rearrest within 10 years; the second is using return to prison within 10 years for a parole or probation violation or a new sentence, a 21 percentage-point gap driven entirely by which event is counted. Now shorten the window: rearrest for the same cohort was 66 percent at 3 years, so trimming the follow-up from ten years to three removes 16 percentage points without changing a single person's behavior. Before comparing any two recidivism figures, pin down three things: the event counted, the length of follow-up, and which released population was tracked. If any of the three differ, the numbers are not comparable.

Key takeaway

Rehabilitation research did not swing from nothing works to everything works; it moved to a conditional answer in which who is treated, what is targeted, how it is delivered, and how well it is implemented determine whether the effect is positive, absent, or negative. Reentry outcomes are reported through a word, recidivism, that hides at least three different measures and follow-up windows, so always ask which event, how long, and which population.

Quick check

3 questions here, of 5 in this lesson’s practice set. Answers stay hidden until you check.

Question 1 of 3foundational

In his 1979 Hofstra Law Review article, what did Robert Martinson do with the conclusion he had published in 1974?

Choose an answer, then check it.
Question 2 of 3intermediate

A state reports that 61 percent of its 2008 prison releases returned to prison within ten years, while a national study reports 82 percent recidivism for the same cohort. What most likely explains the gap?

Choose an answer, then check it.
Question 3 of 3intermediate

A correctional agency has one intensive cognitive-behavioral program with 40 slots. Applying the risk principle as Bonta and Andrews state it, which allocation is best supported by the evidence they report?

Choose an answer, then check it.
Practice all 5

Keep learning

Ready to build on this? Continue to the next lesson.

Practice this lesson
Study tools & related lessonsYou’ll learn to · Common mistakes · Easily confused · Key vocabulary · Related

You’ll learn to

  • Describe what Martinson's 1974 review actually concluded, the caveats he attached to it, and how he revised his position in 1979.
  • Explain the risk, need, and responsivity principles and identify criminogenic needs using the central eight framework.
  • Distinguish rearrest, reconviction, and return to prison as recidivism measures, and explain why follow-up length changes the reported rate.
  • Interpret published effect sizes for correctional programs, including reported nulls, without overstating what a meta-analysis establishes.
  • Analyze a reentry scenario by naming the practical barriers a released person faces and which body of law or evidence governs each.

Common mistakes

  • Citing Martinson for the proposition that nothing works.

    His 1974 article warned that recidivism measures differ, that the research base might be too weak to detect real effects, and that deterrence might be doing unmeasured work. In 1979 he withdrew the conclusion and reported that programs can be beneficial, neutral, or harmful depending on delivery conditions.

  • Treating rearrest, reconviction, and reincarceration as interchangeable measures of recidivism.

    They count different events with different base rates. In the BJS 2008 cohort, 82 percent were rearrested within 10 years while 61 percent returned to prison, and return to prison partly reflects how a state handles supervision violations rather than new offending alone.

  • Assuming that more programming is always better, so low-risk participants should get the intensive version.

    The risk principle says the opposite. Bonta and Andrews report about a 10 percent recidivism difference when treatment goes to higher-risk cases and about 3 percent for lower-risk cases, and they report evidence that intensive services delivered to low-risk people can raise reoffending.

  • Reading a Promising or Effective rating as proof that a local program with the same name will work.

    Ratings attach to evaluated practices and programs, not to labels. Bonta and Andrews report that real-world delivery achieves roughly half the effect of demonstration conditions, so implementation quality, participant selection, and dosage have to be checked locally.

  • Reporting the RAND correctional education finding as a 43 percent drop in recidivism.

    RAND reported 43 percent lower odds of recidivating, which the authors themselves translate into about a 13 percentage-point reduction in risk. Odds ratios and percentage-point changes are different quantities, and the employment findings were weaker still.

Easily confused

Rehabilitation vs. Reentry

Rehabilitation is a claim about changing a person's likelihood of offending, usually through programming. Reentry is the logistical and legal transition out of custody, which happens whether or not any rehabilitative program was delivered.

Rearrest as a recidivism measure vs. Return to prison as a recidivism measure

Rearrest counts a police decision and requires no adjudication, producing the highest rates. Return to prison counts a correctional outcome that may follow either a new sentence or a supervision violation, so it mixes offending with supervision policy.

Martinson's 1974 conclusion vs. The nothing-works slogan

The article concluded that reported rehabilitative efforts had shown no appreciable effect on recidivism, hedged by explicit doubts about measurement and research quality. The slogan dropped the hedges, the retraction, and the 1979 finding that some programs help and some harm.

Criminogenic need vs. Non-criminogenic need

Both can be real needs. Bonta and Andrews report that targeting criminogenic needs is associated with about a 19 percent recidivism difference while targeting non-criminogenic needs is associated with roughly a 1 percent increase, so the distinction is about what the program should aim at, not about what matters to a person.

Key vocabulary

Rehabilitation
The correctional aim of changing a person so that they are less likely to offend again, pursued through programming rather than through incapacitation or deterrence.
Reentry
The transition from confinement back into the community, and the set of practical tasks, supervision conditions, and services that surround it.
Recidivism
A return to criminal behavior after a sanction or intervention, measured in practice by a specified event within a specified follow-up window.
Criminogenic need
A changeable risk factor empirically linked to reoffending, such as procriminal attitudes or substance abuse, as distinct from a need that matters for wellbeing but not for reoffending.
Central eight
Andrews and Bonta's grouping of the major predictors of offending: antisocial personality pattern, procriminal attitudes, social supports for crime, substance abuse, family and marital relationships, school and work, prosocial recreation, and criminal history.
Risk principle
The rule that service intensity should be matched to assessed likelihood of reoffending, with intensive programming reserved for higher-risk cases.
Responsivity principle
The rule that interventions should use cognitive social learning methods and be tuned to the individual's motivation, abilities, and learning style.
Program integrity
The degree to which a program is delivered as designed, at the intended dosage, to the intended participants, with trained and supervised staff.
Follow-up period
The window after release during which a study counts qualifying events; lengthening it raises the reported rate because more people accumulate at least one event.
Conditional release
A release from prison that carries community supervision or other requirements, as opposed to an unconditional release such as an expiration of sentence.

Sources & references

  1. What works? - questions and answers about prison reform, The Public Interest No. 35 (Spring 1974), pp. 22-54 — The Public Interest (National Affairs, Inc.)
  2. New Findings, New Views: A Note of Caution Regarding Sentencing Reform, Hofstra Law Review Vol. 7, Iss. 2, Article 1 (Winter 1979), pp. 243-258 — Hofstra Law Review, Maurice A. Deane School of Law at Hofstra University
  3. Risk-Need-Responsivity Model for Offender Assessment and Rehabilitation (Report 2007-06) — Public Safety Canada
  4. Evaluating the Effectiveness of Correctional Education: A Meta-Analysis of Programs That Provide Education to Incarcerated Adults (RR-266-BJA, 2013) — RAND Corporation, sponsored by the Bureau of Justice Assistance, U.S. Department of Justice
  5. Prisoners in 2023 - Statistical Tables (NCJ 310197) — Bureau of Justice Statistics, U.S. Department of Justice
  6. Jail Inmates in 2023 - Statistical Tables (NCJ 309965) — Bureau of Justice Statistics, U.S. Department of Justice
  7. Recidivism of Prisoners Released in 24 States in 2008: A 10-Year Follow-Up Period (2008-2018) (NCJ 256094) — Bureau of Justice Statistics, U.S. Department of Justice
  8. Recidivism of Prisoners Released in 34 States in 2012: A 5-Year Follow-Up Period (2012-2017) (NCJ 255947) — Bureau of Justice Statistics, U.S. Department of Justice
  9. 2018 Update on Prisoner Recidivism: A 9-Year Follow-Up Period (2005-2014) (NCJ 250975) — Bureau of Justice Statistics, U.S. Department of Justice
  10. Recidivism (topic page) — National Institute of Justice, Office of Justice Programs, U.S. Department of Justice
  11. Practice Profile: Adult Reentry Programs — CrimeSolutions, National Institute of Justice
  12. Practice Profile: Rehabilitation Programs for Adults Convicted of a Crime — CrimeSolutions, National Institute of Justice
  13. Program Profile: Oklahoma's Juvenile Second Chance Act Reentry Program (Tulsa, Okla.) — CrimeSolutions, National Institute of Justice
  14. Second Chance Act of 2007: Community Safety Through Recidivism Prevention, Public Law 110-199 (April 9, 2008) — U.S. Government Publishing Office (govinfo)
  15. 34 U.S.C. 60501 - Purposes; findings — Office of the Law Revision Counsel, U.S. House of Representatives
  16. 42 U.S.C. 1396d - Definitions (medical assistance), including the inmate of a public institution exclusion — Office of the Law Revision Counsel, U.S. House of Representatives
  17. California Penal Code section 3007.05 - Identification cards for persons released from state prison — California Legislative Information (Legislative Counsel of California)
  18. 24 CFR 982.553 - Denial of admission and termination of assistance for criminals and alcohol abusers — U.S. Government Publishing Office / Office of the Federal Register (eCFR)
  19. Enforcement Guidance on the Consideration of Arrest and Conviction Records in Employment Decisions Under Title VII of the Civil Rights Act of 1964 (No. 915.002, April 25, 2012) — U.S. Equal Employment Opportunity Commission
  20. Collateral Consequences: The Crossroads of Punishment, Redemption, and the Effects on Communities (Briefing Report, June 2019) — U.S. Commission on Civil Rights
  21. National Inventory of Collateral Consequences of Conviction (NICCC) — Council of State Governments Justice Center, funded in part by the Bureau of Justice Assistance, U.S. Department of Justice

EliExplains lessons are original prose written from the open, credible references above. See Copyright & Licensing.

Researched 2026-08-19

Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.