监狱装不下人,就重新定义坐牢When Prisons Overflow, Redefine Prison
英国和威尔士今年上半年被召回监狱的人,比被放出去的人还多。4月到6月,12,338人被召回,12,278人被释放(来源:The Guardian)。到6月为止的一年里,52,716人因为违反许可条件被重新关回去,比上一个12个月多了27%。
政府给出的解释很诚实:SDS40提前释放计划把部分犯人的服刑比例从50%砍到40%,于是假释期变长,假释期变长,违规的机会就变多。再加上政策改成多数被召回的人关56天再放出来,而不是关到刑期结束。也就是说,召回人数飙升,一部分是制度设计自己生产出来的。门开得更大,门槛设得更多,两头一起把数字推上去。
这不是在解决监狱危机,是在重新定义什么叫解决。监狱装不下人,就把人提前放出去;放出去的人违反了条件,就再关回来;关回来的人太多,就缩短关押时长再放出去。每一步都是真的,每一个数字都经得起审计,而监狱里的人并没有变少。这套机制在中国科技叙事里我见过:做不出小芯片,就宣布一条新定律,换一把尺子。换尺子不换能力,是同一个动作。只不过这里换的不是纳米,是人。
Howard League的Rob Preece说监狱系统“承受着巨大压力,长期过度拥挤、人手不足、条件恶劣,几乎不可能有效改造人”。这句话是对的,但它说得太客气了。过度拥挤不是天气,是政策选择累积出来的结果。更长刑期制造了拥挤,拥挤催生了提前释放,提前释放催生了更多召回,召回又反过来吃掉本来就不够的空间。这是一个闭环,而闭环里每一环都有人签字。
真正付代价的人,报道里写得很清楚。到2026年9月底为止的12个月里,360人在监狱中死亡,其中66人的死因被记录为“自伤”。到2026年6月底为止的12个月里,监狱记录了68,231起自伤事件,平均每八分钟一起。男监自伤下降了12%,女监下降了3.8%——数字在降,但每八分钟一次意味着它仍然是一种日常。这才是暴力三角里那个差额:监狱系统本可以达到的状态,和它实际给出的状态之间,隔着八分钟一次的自伤和一年66条被记为自伤的死。
还有一个数字值得单独拎出来:281名被判严重罪行的缓刑人员又被控严重犯罪,其中119人是登记在册的性犯罪者,128人是杀人犯,4人是恐怖分子或有恐怖风险。这是连续第三年上升,创下纪录。召回率上升和这个数字之间是什么关系,报道没有给出因果,我不替它补。但可以确定的是,当系统把资源花在反复收放同一批人身上,监管和改造的容量就被挤掉了,而挤掉的部分不会消失,它会在别的地方以别的形式出现。
最值得追问的是“违反许可条件”这个词。它听起来像中性的技术条款,实际上是一个巨大的裁量空间:什么算违规、谁被判定违规、违规之后关多久,全在系统手里。当系统需要腾出空间时,它就把门槛降低;当系统需要显得强硬时,它就把召回数字拿出来说事。同一批人在这两套逻辑之间被来回搬运,而搬运的成本记在他们身上,不记在签字的人身上。
所以这不是一场监狱容量危机,是一场解释权危机。政府把“提前释放”包装成改革,把“召回上升”包装成执行严格,两句话同时成立,因为它们衡量的不是同一个东西。真正被衡量的只有一件事:系统还能不能维持自己看起来在运转。至于每八分钟一次的自伤,和那66个被记录为自伤的死,它们不在任何一句声明里。
In England and Wales, more people were recalled to prison in the first half of this year than were released. Between April and June, 12,338 were recalled and 12,278 released (source: The Guardian). In the year to June, 52,716 people were sent back for breaching licence terms, 27% up on the previous twelve months.
The government's explanation is candid: the SDS40 early release scheme cut the proportion of sentences some prisoners must serve from 50% to 40%, which lengthens the licence period, and a longer licence period means more chances to breach. On top of that, most recalled offenders are now held for 56 days before release rather than until the end of their sentence. So the recall spike is partly manufactured by the policy itself. Open the door wider, add more thresholds, and the numbers climb from both ends.
This is not solving the prison crisis. It is redefining what counts as a solution. Prisons cannot hold everyone, so people are released early; released people breach their terms, so they are recalled; too many are recalled, so the holding period is shortened and they are released again. Every step is real. Every figure survives audit. And the prison population does not shrink. I have seen this mechanism before, in Chinese tech narratives: when you cannot build the small chip, you announce a new law. Changing the ruler is not changing the capacity. Here the ruler is not measured in nanometres. It is measured in people.
Rob Preece of the Howard League says the system is under enormous pressure, with chronic overcrowding, staff shortages and awful conditions that make rehabilitation close to impossible. True, but too polite. Overcrowding is not weather. It is the accumulated result of policy choices. Longer sentences created the crowding, crowding produced early release, early release produced more recalls, and recalls eat the space that was never enough to begin with. It is a closed loop, and at every link someone signed.
The people paying are named in the report. In the twelve months to the end of September 2026, 360 people died in prison, 66 of them in circumstances recorded as self-inflicted. In the twelve months to the end of June 2026, prisons recorded 68,231 incidents of self-harm, one every eight minutes. Self-harm fell 12% in men's prisons and 3.8% in women's prisons. The numbers are falling, and one every eight minutes is still a routine. That is the gap in the Violence Triangle: between what the system could deliver and what it does, there are eight minutes and sixty-six deaths filed under self-inflicted.
One more figure deserves its own paragraph. Of 281 serious offenders on probation charged with a serious further offence, 119 were registered sexual offenders, 128 were murderers, and four were terrorists or posed a terror risk. That is a record, rising for the third successive year. The report does not establish causation between this and the recall rate, and I will not invent one. What is certain is that when a system spends its capacity cycling the same people in and out, supervisory and rehabilitative capacity gets squeezed. Squeezed capacity does not vanish. It reappears elsewhere, in other forms.
The phrase to interrogate is breaching the terms of their licence. It sounds like a neutral technical clause. It is a vast discretionary space: what counts as a breach, who is judged to have breached, how long they are held afterwards. All of it sits with the system. When the system needs space, it lowers the bar. When the system needs to look tough, it waves the recall figures around. The same people are shuttled between two logics, and the cost of the shuttle is booked to them, not to the people signing.
So this is not a prison capacity crisis. It is a crisis of interpretive authority. Early release is packaged as reform, rising recalls are packaged as rigour, and both statements hold because they are not measuring the same thing. The only thing actually being measured is whether the system can keep looking like it works. The self-harm every eight minutes and the sixty-six deaths recorded as self-inflicted do not appear in any statement.