THE RATIO. One judgment, decoded twice: once for anyone, once for the profession.

Headnote

A man accused of strangling another to death in 1999 vanished before he could be tried, while his co-accused stood trial separately and was acquitted. Caught eighteen years later, in 2017, he finally faced trial himself. By then the key eyewitness had died, others had turned hostile, and the one witness still willing to testify could not identify him, in a village with two men of the same name. Courts below convicted him anyway, relying on what the dead witness had said years earlier. The Supreme Court has now acquitted him, for a reason that has nothing to do with whether he did it: nobody ever passed the specific order the law requires before testimony given behind an absconding accused’s back can be used against him once he is finally caught.

Factual Background

On 1 April 1999, witnesses returning from a village fair saw a man sitting on the chest of another, strangling him, while a second man stood by, encouraging the assault and threatening anyone who came close. Frightened, the witnesses fled. The victim died. Two men were eventually charged with his murder. One of them, Mahendra Singh, accused of the actual assault, could not be found. The case against him was separated from that of his co-accused, who was tried and acquitted.

Mahendra Singh remained missing for eighteen years. He was finally arrested on 11 September 2017 and put on trial for the killing. By then, the case had aged badly. The prosecution’s key eyewitness, referred to as PW1, had died in the intervening years. Several other witnesses who had originally supported the prosecution turned hostile at trial, no longer backing their earlier accounts. One eyewitness, PW2, did still testify to having seen the assault, but could not identify Mahendra Singh as the man responsible, a gap made worse by evidence that two different men named Mahendra Singh lived in the same locality where the killing occurred.

Facing this evidentiary gap, the trial court and, on appeal, the High Court, both relied instead on what PW1 had said years earlier, back when the case against Mahendra Singh’s co-accused was being heard, at a time when Mahendra Singh himself was still absconding. That earlier account had clearly named Mahendra Singh as the killer. On the strength of it, he was convicted. He appealed to the Supreme Court.

The Question for Determination

Indian criminal procedure allows a court to record a witness’s evidence even in the absence of an accused who is deliberately evading arrest, so the evidence is not lost to the passage of time. Section 299 of the Code of Criminal Procedure governs this, but does so in two distinct stages. Before any such evidence can be recorded at all, the court must first be satisfied of two things together, that the accused is absconding, and that there is no immediate prospect of arresting him. Only later, if the accused is eventually caught and that same testimony is to be used against him without giving him any chance to cross-examine the witness, must the court separately find that the witness has since died, become incapable of testifying, or cannot be traced or produced without unreasonable delay or expense.

The question in Mahendra Singh’s case was narrow but decisive: had that first, foundational order ever actually been made, back in 1999, when PW1’s evidence was recorded while Mahendra Singh was absconding and his co-accused was being tried. If it had not, the later reliance on PW1’s testimony against Mahendra Singh, once he was finally caught, would have no proper legal foundation at all.

The Court’s Reasoning

The Court traced how this two part structure has been consistently understood. In Nirmal Singh v State of Haryana, it had already established that Section 299 splits into a first part governing when evidence may be recorded in an accused’s absence, and a second governing when that evidence may later be used without cross-examination. In Jayendra Vishnu Thakur v State of Maharashtra, a case under anti-terrorism legislation where an order had been passed despite the investigating officer himself knowing the accused had already been arrested elsewhere, the Court went further: the two conditions in the first part, abscondence and the absence of any immediate prospect of arrest, must be proved conjunctively, with a definite, reasoned finding, not simply assumed or asserted. Sukhpal Singh v NCT of Delhi, a case with a strikingly similar fact pattern to this one, a killer who fled and was caught two decades later, reaffirmed the same structure and upheld reliance on an earlier witness statement only because the required order had, in that case, actually been passed at the time.

Applying that consistent line of authority, the Court found the fatal gap in Mahendra Singh’s case. No order under section 299 had ever been passed in 1999, either when the case was committed for trial or when the trial against his co-accused began. Mahendra Singh’s flight, and the impossibility of arresting him at the time, had never been proved to any court’s satisfaction, because no court was ever asked to make that finding. Without it, PW1’s testimony, however clearly it may have named Mahendra Singh at the time, could not be treated as evidence properly preserved against him under section 299. It was, in effect, testimony given in someone else’s trial, never validly connected to his.

With that foundation absent, and with the trial’s own live evidence, PW2’s inability to identify him amid confusion over two men sharing his name, and the other witnesses having turned hostile, insufficient on its own to convict, the prosecution’s case failed entirely. The Court allowed the appeal, acquitted Mahendra Singh, and ordered his release, or the cancellation of his bail bonds if he had already been released pending appeal.

Critical Assessment

Section 299 rests on the principle that a man who deliberately flees to frustrate his own trial cannot later complain that he was denied the chance to cross-examine a witness, since nobody may claim the benefit of a wrong he committed himself. This case shows the limit of that principle rather than its application. Mahendra Singh almost certainly was absconding in 1999, and there was almost certainly no immediate prospect of arresting him at the time, exactly the situation section 299 exists to address. What defeated the prosecution was not any wrong committed by him beyond the flight itself, but a wrong committed by the system meant to respond to it: nobody asked a court to make the finding the law requires, at the moment it needed to be made. The accused ends up benefiting not from his own evasion as such, but from an administrative gap in how that evasion was handled at the time.

That gap carries a clear lesson for how cases with multiple accused and one absconder should actually be run. Where a trial proceeds against one accused while a co-accused remains at large, there is an obvious pull toward simply getting on with the case in front of the court, since evidence is being recorded and a trial is underway regardless. This case shows why that instinct is dangerous. Preserving evidence against an absent co-accused is not automatic just because a related trial is happening. It requires its own, separate, formally reasoned order, sought and passed at the time, precisely because by the time anyone thinks to check whether it was done, the absconder may already have been caught, tried, and convicted on evidence that never actually applied to him. The cost of skipping that step is not felt for years, sometimes decades, which is exactly what makes it easy to skip.

The facts here also show why the rule is not empty formalism. Two men shared the name Mahendra Singh in the same locality, a detail that surfaced only because the surviving witness was cross-examined and could not make a positive identification. Had PW1’s old testimony been treated as automatically usable, that risk of misidentification would never have been tested at all. A rule that forces a fresh, present day check on whether the evidence against a particular defendant still holds up is doing real protective work, not merely enforcing a technicality for its own sake.

Implications

If you’re not a lawyer: this case is a stark illustration of how a criminal case can turn on something that happened, or failed to happen, in a courtroom decades before the trial that actually decides guilt or innocence. If you are ever a witness in a case where a co-accused has fled, your evidence is not automatically preserved for use against that person later, a specific court order has to be sought and made at the time, and if it is not, that evidence may become unusable no matter how clearly you remember what you saw. It is also a reminder that eyewitness identification can fail in very ordinary ways, including something as simple as two people sharing the same name in the same village.

For the profession: this confirms, in a fact pattern almost identical to Sukhpal Singh, that the two jurisdictional facts under section 299, abscondence and the absence of any immediate prospect of arrest, must be established by a definite, reasoned order at the time evidence is recorded, not inferred later from the surrounding record. Prosecutors handling any matter with an absconding co-accused should treat obtaining a proper section 299 order as a non-negotiable, contemporaneous step, independent of how the trial against a present co-accused is proceeding, since the omission is invisible until the absconder is eventually caught, by which point it is generally too late to cure. Defence counsel facing reliance on old testimony recorded during a co-accused’s trial should check, as a first move, whether any section 299 order was ever actually passed against their own client specifically, since its absence is a complete answer regardless of how compelling the earlier testimony reads on paper.

Case Details

  • Citation: 2026 INSC 775, Criminal Appeal No. 3566 of 2026, arising out of SLP(Crl) No. 6741 of 2026
  • Court: Supreme Court of India, Criminal Appellate Jurisdiction. J.B. Pardiwala J and K. Vinod Chandran J
  • Decided: 31 July 2026
  • Key precedents applied: Nirmal Singh v State of Haryana, (2000) 4 SCC 41 · Jayendra Vishnu Thakur v State of Maharashtra, (2009) 7 SCC 104 · Sukhpal Singh v NCT of Delhi, 2024 SCC OnLine SC 800
  • Governing provisions: Section 299, Code of Criminal Procedure, 1973. Section 33, Indian Evidence Act, 1872
  • Key paragraphs: [3]-[9] (the two part structure of Section 299, and the conjunctive jurisdictional facts) · [11]-[12] (the evidentiary gap at trial: no identification, hostile witnesses) · [13]-[14] (no Section 299 order ever passed in 1999; acquittal follows)