Prison Types, Classification of Prisoners, and Institutional Hierarchy

Custody, in its narrowest sense, means only that a person is physically confined and prevented from leaving. Institutional treatment is a broader idea: it names everything a prison administration does with a confined person once custody is established, including how that person is housed, what work or education they are assigned, what disciplinary code governs their conduct, and what preparation is made for their eventual release.

The distinction matters because a jail can achieve custody with almost no institutional treatment at all, simply by locking people in cells, and historically many did exactly that.

The shift toward treating imprisonment as a managed process rather than bare confinement tracks the broader move from purely retributive punishment toward reformative goals, discussed in the chapter on aims of punishment. Once reform becomes a stated goal, the administration needs tools to decide who needs what kind of intervention, and that need is what makes classification necessary rather than optional.

Saleilles's individualisation principle supplies the reasoning: two people convicted of the same offence can present entirely different levels of risk and entirely different needs, and treating both identically wastes resources on the low-risk case while under-managing the high-risk one.

Classification operationalises this for an administration that cannot design a bespoke regime for every prisoner, sorting prisoners into a manageable number of bands, each with a pre-designed regime, so the system approximates an individual response within the practical limits of a large institution.

Classification also protects prisoners from each other. Mixing an undertrial held on a minor charge with a convicted repeat violent offender exposes the former to exploitation and coercion, a risk correctional literature associates with prisonisation, the absorption of the norms of the criminal subculture.

In India, this apparatus sits inside state prison manuals built on the Model Prison Manual and, since 2023, the Model Prisons and Correctional Services Act, a template law states can adopt because prisons are a State List subject; the 2023 model act provides for a Classification and Security Review Board in every prison.

Most classification systems, whatever their jurisdiction, draw on a common set of criteria. Offence type separates violent from non-violent offenders and often further separates categories such as sexual offences, which typically require distinct housing and programming considerations.

Sentence length is used both because longer sentences are treated as a proxy for offence seriousness and because a prisoner facing a decade or more in custody needs a different institutional environment, with more programming and a longer planning horizon, than one facing a few months.

Age is a near-universal separator: juveniles are held apart from adults as a matter of law rather than administrative discretion in essentially every jurisdiction discussed in this chapter, reflecting a distinct legal framework for juvenile justice.

Sex is likewise a mandatory separator, with women's facilities run as distinct institutions or wings rather than integrated into general male facilities, partly for safety and partly because women's imprisonment patterns, typically shorter sentences and higher rates of primary caregiving responsibility, call for different programming.

Risk assessment is the most technical criterion, shaped in India by the Model Prisons and Correctional Services Act's proposed Classification and Security Review Board, which is directed to weigh escape risk, security requirements, and physical and mental health needs alongside the more static factors of age, sex, and sentence length.

Modern risk assessment increasingly uses structured instruments scoring criminal history and disciplinary record rather than relying solely on an officer's judgement, though structured tools remain only partially adopted across India's prison system.

A final criterion, often underweighted in older manuals but emphasised in reform proposals, is the distinction between undertrials awaiting trial and convicted prisoners. In India, undertrials form a large majority of the total prison population, and separating them from convicts is meant to prevent people who are legally presumed innocent from being managed under the same security regime as sentenced offenders, though overcrowding often erodes this separation in practice, a problem returned to later in this chapter.

Once prisoners are classified, that classification has to translate into a physical placement, and most systems use a security-level typology to make the translation legible. The generic three-tier model, maximum, medium, and minimum security, describes facilities by their perimeter security, internal movement controls, and staff-to-prisoner ratio rather than by the prisoners housed there directly, though in practice the two track closely.

The Federal Bureau of Prisons in the United States refines this into five institution security levels, minimum, low, medium, high, and administrative, the last reserved for special-purpose facilities such as pretrial detention centres and medical units rather than a point on the ordinary risk spectrum. Placement uses a points-based custody classification scoring offence severity, criminal history, sentence length, and escape history, with factors such as a history of serious violence forcing a higher security placement regardless of the numeric score.

England and Wales uses a different mechanic for the same purpose: rather than grading institutions, His Majesty's Prison and Probation Service assigns each sentenced adult male prisoner a personal security category, A through D. Category A prisoners are those whose escape would be highly dangerous to the public, the police, or the security of the state, and are held under the highest security regardless of which prison they are in.

Category D prisoners are trusted to serve their sentence in open conditions, so the category attaches to the person and travels with them between institutions, unlike the American model where the security level attaches primarily to the facility.

Both systems illustrate a common tension: security-level typologies must be coarse enough to be administratively workable across large prisoner populations, yet fine enough to avoid either warehousing low-risk prisoners under unnecessarily restrictive conditions or placing high-risk prisoners in facilities that cannot contain them. Reclassification review, addressed later in this chapter, exists to correct initial placements that turn out to be miscalibrated.

India's prison system does not attach a security-level label to a named institution type the way the American or British systems do. Instead, the hierarchy is organised by administrative tier and catchment area, with security layered on top rather than replacing it.

According to the National Crime Records Bureau's Prison Statistics India report, the country's 1,319 prisons break down as 564 sub-jails, 424 district jails, 148 central jails, 88 open jails, 41 special jails, 32 women's jails, 19 borstal schools, and a small number of other facility types.

Central jails sit at the top of the ordinary hierarchy, serving a wide catchment, holding the largest prisoner populations, and reserved primarily for prisoners sentenced to more than two years, a threshold that functions as India's rough equivalent of a sentence-length criterion; they also carry the most developed vocational and educational programmes.

District jails sit in the middle tier, one per district in most states, holding a mixed population of undertrials and shorter-term convicts, and where a district has no central jail its district jail may hold longer-term convicts too. Sub-jails form the base, small lock-ups at the sub-division or taluka level used mainly for short remand and transit custody.

This tiered structure traces back to the Prisons Act, 1894, refined through state manuals and the Model Prison Manual, with the 2023 Model Prisons and Correctional Services Act intended to modernise the base rather than replace the three-tier structure.

Alongside this vertical spine sit special institution types that classify by prisoner category rather than catchment area. Women's jails, 32 of them, hold female prisoners separately from the general male population as a legal requirement rather than a discretionary choice.

Borstal schools, 19 of them, are reformatory institutions for young offenders kept apart from adult jails; the model itself originated in England, named after the village of Borstal in Kent, and was adopted into Indian legislation through state-level Borstal Schools Acts well before the current reform cycle.

Open institutions, 88 of them, hold prisoners assessed as low risk under minimal perimeter security, often on agricultural or industrial work with day release; Rajasthan has run open camps for decades to prepare long-term convicts for release under reduced supervision.

Special jails, a further 41 institutions, cover facilities with a narrow purpose, such as high-security prisons for prisoners linked to organised crime, sitting at the opposite end of the risk spectrum from open institutions. A women's jail or an open jail is not itself a rung on the central-district-sub-jail ladder; it is a separate classification dimension layered onto it.

The United States runs a hierarchy shaped by its federal structure rather than a single national tier system. Federal prisoners are held in Bureau of Prisons facilities graded by the five-level security typology described earlier, while state prisoners are held in a separate state-run system whose naming conventions, correctional facility, state penitentiary, and similar labels, vary by state and do not map cleanly onto the federal levels.

Below both sits a third, largely separate tier of county and city jails, which hold pretrial detainees and short-sentence offenders and are typically run by a county sheriff's department, a structural split with no exact Indian counterpart because Indian sub-jails remain part of the same state prison department as central and district jails.

England and Wales, by contrast, runs a single national prison service, so there is no federal-state split; the hierarchy instead runs through the personal category system, A to D, layered onto a set of local, training, and open prisons.

Local prisons serve a court catchment and hold a mixed population including new receptions and remand prisoners, functioning somewhat like India's district jails, while training prisons hold longer-term convicted prisoners assigned by category and offer programming comparable to an Indian central jail's role, and open prisons hold Category D prisoners under conditions comparable to India's open jails.

Despite the structural differences, three functions recur across all three systems: a facility or personal grading tied to escape and violence risk, a separate track for prisoners serving longer sentences that concentrates reformative resources, and a distinct low-security track for prisoners nearing release.

What differs is whether the grading attaches to the institution, as in the American federal model and, loosely, the Indian tier system, or to the individual prisoner as they move between institutions, as in the English category system.

This comparison matters for reform debates in India, because proposals to introduce a more formal, points-based custody classification, closer to the American model, have to reckon with the fact that India's tier structure is built around fixed institutions serving fixed catchment areas rather than around individually graded prisoners who can be moved to whichever institution matches their score, a structural constraint the Model Prisons and Correctional Services Act's Classification and Security Review Board framework does not by itself remove.

Classification is not a single decision made once at the prison gate; it is a process with distinct stages. Reception covers identity verification, a basic medical screen, and recording of the charge or sentence, catching immediate risks, self-harm risk, communicable disease, or a known enemy relationship with another prisoner, before entry to general population.

Diagnostic assessment follows and is meant to be more thorough, examining criminal history, education, and psychological indicators to inform both security placement and a programming plan.

The Model Prisons and Correctional Services Act's proposed Classification and Security Review Board is meant to formalise this as an ongoing function, with authority to revise a prisoner's classification periodically rather than leaving the initial placement fixed for the sentence.

In practice, the extent of formal diagnostic assessment varies considerably across Indian states and often depends on whether a jail has dedicated welfare and psychology staff, a resource that is unevenly distributed; India's National Human Rights Commission and successive Prison Statistics India volumes have repeatedly flagged staff shortages as a constraint on how thoroughly classification criteria are actually applied, as distinct from how they are written in state manuals.

Every classification system described in this chapter assumes a facility has enough physical capacity, in the right security band, to implement its placement decisions. India's Prison Statistics India data has for years recorded national occupancy above rated capacity, concentrated among undertrials.

Overcrowding erodes classification in a predictable way: when a district jail is at or above capacity, the pressure to accommodate new admissions pushes prisoners into whatever space is available rather than the space their classification would ideally assign, so undertrials end up housed alongside convicts and prisoners of differing risk levels share barracks designed for a smaller, more uniform population.

The consequences are not merely administrative. Mixing risk levels under overcrowded conditions is associated with higher rates of violence, and it undermines the reformative half of institutional treatment, since programming concentrated in central jails cannot scale to a population well beyond design capacity.

The 2023 model act addresses overcrowding only indirectly, encouraging alternatives such as parole, furlough, and open institutions rather than expanding physical capacity, so whether classification functions as designed in a given facility depends on how close its actual population is to its rated capacity.