Registration is the beginning of a chain: investigation, a report to the court, a prosecutor, a trial, and, for most people caught in it, a cell while they wait. Each handoff is a place where a case can quietly die.
Public argument about criminal justice in Pakistan tends to arrive at the judge. The judge acquitted, the judge delayed, the judge should have known. It is the wrong end of the pipe. A judge decides on the file placed before them, and almost everything in that file was fixed months earlier by people the public never names: the officer who did or did not visit the scene, the prosecutor who did or did not send the report back, the constable who did or did not produce the witness. The chain from an FIR to a verdict has five handoffs. Most cases lose their outcome at one of the first two.
Once a cognizable offence is registered, the police may investigate without a magistrate's order. Under section 157 of the Code of Criminal Procedure, 1898, the officer in charge sends a report to the magistrate and proceeds to the spot. What the investigating officer does next — the scene, the seizures, the medical examination, the people spoken to — is the evidence. Nothing later manufactures it.
Since 2002 the law has tried to make this a specialism rather than a chore. The Police Order, 2002 organised the force into functional branches, investigation and watch and ward among them, and put a separate head of investigation in each district who is not below the rank of Superintendent, reports to their own hierarchy, and whose process the District Police Officer is expressly forbidden to interfere with. Provinces have since gone their own ways — Khyber Pakhtunkhwa legislated its own Police Act in 2017 — but the design idea is common: the officer who keeps order should not be the officer who builds the case.
The design assumes capacity that is not evenly there. The National Commission for Human Rights, in its report on police reforms, called the separation a step in the right direction while describing investigative technique as outmoded and the branch as chronically under-resourced and under-trained. That is not a small observation. Statements the investigating officer records under section 161 are not signed by the witness and, under section 162, may generally be used at trial only to contradict the person who made them. So a case that rests on what people said to the police, rather than on what can be proved independently, is fragile by construction.
The document that turns an investigation into a case is the report under section 173, known everywhere as the challan. On completing the investigation the officer in charge forwards it, through the Public Prosecutor, to the magistrate empowered to take cognizance — naming the parties, setting out the offence, listing the witnesses, and stating whether the accused is in custody or on bond. The Code also sets a clock: where the investigation is not completed within fourteen days of the FIR, an interim report must be submitted within three days after that, so the court can begin.
Two features decide a great deal. First, the challan, not the FIR, is the operative document; it can name people the FIR did not and drop people it did. Second, the court is not obliged to agree with it. A magistrate may take cognizance on the material, or decline to. What the court cannot do is supply what the file lacks. If the site was never examined and the recovery was never linked, no amount of judicial attention creates the evidence.
Until the 2000s the person who conducted the prosecution was, in practice, downstream of the police in every way that mattered. All four provinces then legislated separate prosecution services: Balochistan in 2003, Khyber Pakhtunkhwa in 2005, Punjab in 2006 and Sindh in 2009. Each Act creates a Prosecutor General, a cadre of prosecutors, and the stated aim of an independent and effective service so that the decision to prosecute is separated from the decision to investigate.
The operative power is small and important. Under section 9 of the Punjab Act, a prosecutor who finds a police report defective must return it to the officer within three days for the defects to be removed, and otherwise files it in court; Sindh's Act carries the same three-day scrutiny. That is the system's quality gate. What the statutes do not give the prosecutor is any authority to direct an investigation while it is running — a gap the Research Society of International Law identifies directly in its survey of the prosecution's role. The prosecutor can refuse to accept a bad file. The prosecutor cannot make a good one.
The prosecutor can refuse a bad file. The prosecutor cannot make a good one. That is the whole of the design problem in one sentence.
At trial the court frames a charge, the prosecution leads its evidence and the defence cross-examines, the accused is examined under section 342, and judgment follows. The point every practitioner makes is that the trial tests the file rather than the truth. A witness who does not appear, or appears and resiles, ends the case regardless of what happened. Writing in The Express Tribune on 31 August 2025, Manzar Zaidi, returning to his own earlier study of terrorism prosecutions, described the recurring pattern in anti-terrorism courts: cases padded with weak witnesses, witnesses who turn hostile out of fear, an overwhelming reliance on eyewitness testimony with forensic evidence used only as a supplement, and prosecutors who present a file without challenging its defects.
Meanwhile the accused waits. Section 497 allows bail in non-bailable offences and, by its provisos, requires release where trial has not concluded within one year of continuous detention for offences not punishable by death, or two years where death is a possible sentence, provided the delay is not the accused's own doing — with exceptions for previous convicts, those the court holds to be hardened, desperate or dangerous, and terrorism offences carrying death or life. Those periods are the law's own admission of how long a case can take.
| Stage | Instrument | Who decides | How a case dies here |
|---|---|---|---|
| Registration | Section 154, CrPC | Officer in charge of the police station | Wrong sections; delay; no copy to the informant |
| Investigation | Sections 157 and 161, CrPC | The investigating officer | Scene not worked; evidence not linked; witnesses only on paper |
| Report | Section 173, CrPC | Officer in charge, through the prosecutor | Incomplete challan; the interim report becomes the case |
| Scrutiny | Provincial prosecution Acts | The prosecutor | A defective file passed on rather than returned |
| Trial | Charge, evidence, section 342 | The trial court | Witnesses absent or hostile; nothing to convict on |
| Custody | Section 497, CrPC | The court, on bail | The wait becomes the punishment |
The clearest measure of a criminal justice system is not how many people it convicts but who is inside its prisons. The National Commission for Human Rights, with the National Academy of Prisons Administration and Justice Project Pakistan, published national prison data in January 2025 covering July and August 2024. It counted 102,026 prisoners against an authorised capacity of 65,811 — about 152 per cent occupancy — of whom 74,918, or 73.41 per cent, were under trial rather than convicted. That share has risen from 66 per cent in 2017.
Read that number carefully, because it is the whole argument. Nearly three in four people held in Pakistani prisons have been convicted of nothing. They are there because a case is pending, and the case is pending because the chain above is slow at every joint. The Law and Justice Commission of Pakistan's judicial statistics for 2024 record more than 2.3 million cases pending nationally on 31 December 2024, roughly four-fifths of them in the district judiciary — the courts where criminal trials actually happen.
Nearly three in four people in Pakistan's prisons have been convicted of nothing. That is not a sentencing statistic. It is a measure of the queue.
The reason to trace the chain is that it relocates the argument. Prosecutions succeed or fail on the quality of what is collected in the first days and on whether anybody with authority refuses to pass a weak file forward. Both of those happen long before a judge sees anything. Reform arguments that stop at the courts are aimed at the last joint in a chain of five.
None of this needs a new theory of the state. It needs the ordinary observation that a verdict is assembled by many hands, most of them not in a courtroom, and that the person who pays for a weak file is usually the one sitting in a cell waiting for someone to finish it.