Lokesh Bhardwaj Law Offices Lokesh Bhardwaj Law Offices ← Articles
Lokesh Bhardwaj Law Offices · Advocates and Legal Consultants, New Delhi · lokeshbhardwajlawoffices.com
Constitutional Law · Essay

Why should a judge recuse? And who should decide?

The case for a codified SOP and a Reference Bench for judicial recusal in India.

21 April 2026 17 minute read
Last reviewed: 17/10/2026
the reference bench, absent nemo judex in causa sua
The judge, the gavel, and the closed circle the principle is meant to break.

There is one subject on which the Indian Bar maintains an almost professional silence. It is the subject of judicial recusal. We discuss it sparingly in chambers. We allude to it cautiously when circumstances force the question into the open. We almost never commit our views to writing, for reasons that are obvious to anyone who practises in the constitutional courts.

The cost of that silence, however, is one we ought no longer to be willing to bear, because the procedure by which recusal is sought, heard and decided in this country rests on an unresolved contradiction at its very foundation.

Two principles hold up the architecture of natural justice. Audi alteram partem, that no person be condemned unheard. Nemo judex in causa sua, older and more uncompromising, that no person sit in judgment over a cause in which he has an interest. It is the second principle that concerns me here, because it is the principle our recusal procedure quietly violates while purporting to enforce.

When a litigant rises and asks a judge to recuse, the application is, by present convention, heard and decided by the very judge whose recusal is sought. The judge listens to the question raised against himself. The judge writes the order on whether the question is well founded. The judge resumes the seat regardless. The first principle of natural justice is, in that moment, suspended for the very application that invokes it. I have not, in my years at the Bar, found myself reconciled to this arrangement. I do not believe I am alone.

I. What recusal actually is

Recusal is the act by which a judge withdraws from hearing a matter on the ground that there exists, or appears to exist, a circumstance which would impair the impartial discharge of judicial function. The recognised grounds are familiar to every practitioner of any standing: pecuniary interest, however slight; prior involvement in the matter as counsel; family or social proximity to a party; prior expression of opinion on the very question that now falls for decision; and, in the residuary category, any other circumstance which would lead a fair-minded and informed observer to apprehend a real possibility of bias.

Recusal is not a confession of bias. It is not an indictment of the bench. It is, rightly understood, an affirmation of judicial integrity, because it is the moment at which the judge places the institution above the case in his hands. The cleanest justice in any system is delivered by a judge who knows when to step aside.

II. The foundations of the principle

The rule against bias is older than any constitution and stands in no need of statutory affirmation. A person who has an interest in a matter cannot decide it justly. The proposition is self-evident in every other walk of life. Why it should ever stand in doubt when applied to the bench is the question that has long troubled me.

Lord Hewart C.J. articulated the principle in R v. Sussex Justices, ex parte McCarthy [1924] 1 KB 256 in a sentence that no subsequent reformulation has improved upon:

Justice should not only be done, but should manifestly and undoubtedly be seen to be done.

The integrity of the bench is not a matter of what the bench believes about itself. It is a matter of what the litigant, and through him the public, perceives. Strip away the perception and the judgment loses its moral authority, even where, as a matter of fact, the bench is wholly impartial.

Our own Supreme Court has affirmed this in successive decisions. The Constitution Bench in A.K. Kraipak v. Union of India, (1969) 2 SCC 262, held the rule against bias to be a facet of natural justice applicable across the spectrum of decision-making, judicial and administrative alike. Manak Lal v. Dr. Prem Chand Singhvi, AIR 1957 SC 425, established that the test is not actual bias but the reasonable apprehension of it. Ranjit Thakur v. Union of India, (1987) 4 SCC 611, refined the formulation in language to which I find myself returning often: the question is whether a reasonable person, viewing the matter realistically, would think it likely or probable that the judge would have been biased.

The test is not what the judge thinks of his own conscience. The test belongs to the eye of the litigant, and ultimately, to the eye of the public.

III. The three duties of the litigant

Before any litigant approaches the bar with a recusal application, three duties must be discharged.

The first is the duty of candour. The application must rest on facts that are identifiable and articulable. It cannot rest on suspicion. It cannot rest on dissatisfaction with an interlocutory order. It cannot rest on the corridor whisper that the bench is unfavourable. There must be a fact, or a constellation of facts, which a reasonable observer would regard as raising a real possibility of bias.

The second is the duty of restraint. A recusal application is not, and must never be allowed to become, an attack on the judge. It is a respectful invitation to the judge to consider whether something exists which, viewed from beyond the bench, would impair confidence in the eventual judgment. The language of the application matters. Tone matters. The Indian Bar has, on more than one occasion, paid a heavy institutional price for confusing a legitimate recusal application with a personal accusation. Those episodes are object lessons that the profession ought never to forget.

The third is the duty of timing. The application must be moved at the earliest opportunity. A litigant who knows of a ground of bias and yet proceeds with the hearing waives it. He cannot keep it in reserve, to be deployed only when the order is found to go against him. Manak Lal settled the principle of waiver in 1957, and nothing in subsequent authority has weakened it.

IV. The paradox at the heart of the procedure

A recusal application, by present convention in India, is heard and decided by the very judge whose recusal is sought. The judge listens. The judge considers. The judge rules. If the judge declines, the matter proceeds before the same bench. The litigant has no immediate, separate appeal. His remedy lies somewhere down the road, in the appellate framework, which may, eventually, set aside the final order on the ground of bias.

The justifications offered for this convention are pragmatic. The judge alone, it is said, knows the contents of his own conscience. The judge alone can speak with authority to his prior dealings. To allow another judge to decide would invite forum shopping, disrupt the roster, weaken the bench, and erode the dignity of the office. Each of these justifications carries weight. None of them, in my respectful submission, disposes of the question.

The question is this. The principle that animates recusal is nemo judex in causa sua. The principle that the recusal procedure violates is nemo judex in causa sua. We tell judges to recuse because no person should judge his own cause. We then ask the same judge to decide whether to recuse, which is, on any honest reading, the very act of judging his own cause.

The procedure asks the principle to enforce its own breach.

This is not a difficulty that resolves itself by silence. It can only be resolved by reform.

V. The part-heard lock

There is a feature of our procedure that compounds the paradox in a way few practitioners pause to consider, but every practitioner has experienced.

When a judge has substantively engaged with a matter, even at an interlocutory stage, the matter becomes part-heard before that bench. Part-heard matters, by long convention of our courts, do not circulate with the determination. They do not move with the roster. When the determination is reassigned, when the bench is reconstituted at the beginning of a new term, when the judge passes from one division of work to another, the part-heard matter follows the judge. The judge's seizin, once attached, becomes institutional. It outlives the roster that produced it.

The consequence deserves to be stated plainly. In any system without a part-heard convention, the natural circulation of judicial work over months and terms would, in itself, dilute concerns of bias. A litigant uneasy with the conduct of one bench would, in the ordinary course of listing, find his matter placed before another in the next term. Time and roster would accomplish, quietly and without confrontation, what no recusal application could ever accomplish openly.

Our part-heard convention removes that quiet remedy. Once a judge takes seizin of a matter, the matter remains his until final disposal, irrespective of how the roster otherwise rearranges itself around him. When the same judge then hears and declines a recusal application addressed to himself, the litigant is left without recourse, not only against the judge's own view of his impartiality, but against the procedural architecture that locks the matter to the judge for the duration of its life. The judge before whom the recusal application is moved is, in a real and practical sense, the judge for the life of the case.

This is not a criticism of the part-heard convention as such. The convention serves obvious purposes: continuity, judicial economy, avoidance of duplication, the conservation of institutional memory acquired in long-running matters. But it does mean that, in the Indian context, a judge who declines to recuse is not merely declining to step away from a hearing. He is declaring that he, and he alone, will decide the fate of this case, regardless of how the rest of the institution moves around it. That declaration may, in any given matter, be entirely justified. But it is a declaration of substantial weight, and the procedure by which it is made deserves to be more carefully constructed than it presently is.

VI. Should the same judge decide?

I do not pretend the answer is straightforward. There are weighty considerations on both sides.

The argument for the same-judge model is genuine. The judge alone has direct access to his own state of mind, his own prior dealings, his own conscience. A system in which any judge's recusal could be procured merely by reference to another judge would weaken benches, encourage tactical filings, and inflict avoidable injury on the institution. It would also, in time, transform the recusal application from a serious matter of conscience into a routine instrument of inconvenience.

The argument for the different-judge model is, in my respectful view, the stronger one. The matter is not, in truth, about the judge's state of mind at all. It is about the appearance of justice from where the litigant stands. That appearance is best assessed by a person other than the one whose appearance is in question. To delegate the assessment to the same judge is to make the test of appearance turn on the very perception we are attempting to test.

A balanced position, neither extreme, would be this. The application ought to be moved before the same judge in the first instance, who should hear both sides and decide by a reasoned order. If the application is declined, the litigant ought to have an immediate, time-bound right of reference to the Chief Justice of the court concerned, who would either place the recusal question alone, and not the merits, before a designated bench for limited determination, or constitute a standing recusal panel for that limited purpose.

Such a procedure would preserve the judge's primacy. It would protect the roster. It would honour institutional dignity. It would also, at long last, acknowledge what the principle has been telling us all along: that the question of appearance is a question for someone other than the person whose appearance is questioned.

VII. The SOPs we already have, scattered across the reports

We are not without guidance. Over a century and a half of decided cases, courts in England and India have laid down what amounts to a body of standard operating procedures, scattered across the law reports. The work of consolidation has not been done. The materials, however, are abundant and largely consistent.

From England: Dimes v. Grand Junction Canal Proprietors (1852) 3 HLC 759 established the rule of automatic disqualification where a judge has a pecuniary interest, however small. R v. Bow Street Metropolitan Stipendiary Magistrate, ex parte Pinochet Ugarte (No. 2) [2000] 1 AC 119 extended the principle of automatic disqualification to non-pecuniary interests, where the judge has a sufficiently close connection with a cause being promoted by a party. Locabail (UK) Ltd v. Bayfield Properties Ltd [2000] QB 451 supplied the now-standard taxonomy of recusal grounds and clarified what does, and what does not, amount to apparent bias. Porter v. Magill [2001] UKHL 67 settled the modern formulation: whether a fair-minded and informed observer, having considered the facts, would conclude that there was a real possibility of bias.

From India: Manak Lal (1957), A.K. Kraipak (1969), Ranjit Thakur (1987), State of West Bengal v. Shivananda Pathak, (1998) 5 SCC 513, Supreme Court Advocates-on-Record Association v. Union of India, (2016) 5 SCC 808, and Indore Development Authority v. Manoharlal, (2020) 8 SCC 129, between them, supply the present Indian framework. The test is reasonable apprehension viewed from the litigant's vantage. The application must be moved promptly. The judge must record reasons whether the application is accepted or declined. The order is amenable to challenge in the appellate framework. Tactical or speculative applications will be deprecated.

To this body of judicial work must be added the Bangalore Principles of Judicial Conduct, 2002, adopted by the United Nations and treated as a benchmark by the Indian judiciary, and the Restatement of Values of Judicial Life, resolved by the Supreme Court of India in 1997. Both supply the ethical scaffolding within which the case law operates.

VIII. What is still missing

What we lack, despite the abundance, is a single, codified, time-bound procedural framework. The fragments are present in the case law. They have never been gathered into a published SOP that a litigant, a counsel, or a registrar can hold up and follow.

A consolidated Indian SOP on recusal ought, at the minimum, to address six matters.

First, the categories of mandatory disclosure that every judge must make at the threshold of a hearing, drawing upon the model of the Fifth and Seventh Schedules to the Arbitration and Conciliation Act, 1996, and upon the Bangalore Principles. If we require disclosure of arbitrators, the case for requiring it of judges is, a fortiori, stronger.

Second, the form, content and timing requirements of a recusal application, including the obligation of candour and the consequences of waiver.

Third, the procedure for hearing the application: whether in chamber or in open court, the right of all parties to be heard, and the mandatory requirement of a reasoned order.

Fourth, the remedy where recusal is declined, including a time-bound right of reference to the Chief Justice of the court concerned, and the constitution of a designated recusal-deciding panel for that limited purpose. This is the structural reform that the principle most urgently demands.

Fifth, protection of the judge against tactical or frivolous applications, including the power to award costs and to record observations on the conduct of counsel responsible.

Sixth, the publication of recusal orders, suitably redacted, so that an accessible body of recusal jurisprudence may grow over time. Today, recusal is among the least transparent areas of judicial decision-making, precisely because the orders rarely enter the published record. That fact alone is a measure of how far the principle still has to travel before it is fully at home in our system.

IX. In closing

Recusal is not a sign of weakness in the bench. It is a sign of strength in the institution. A judge who recuses preserves not only the litigant's faith in the judgment, but the public's faith in the judiciary. A system that allows the question of recusal to be raised, heard, and decided fairly, openly, and with the dignity it deserves, strengthens itself with every recusal it administers correctly, and weakens itself with every recusal it administers carelessly.

I write as a practising advocate, not as a critic of the bench. I have the deepest respect for the institution before which I appear daily. That respect is precisely why I believe the recusal procedure deserves a more honest architecture than it presently has.

Nemo judex in causa sua is not a slogan to be inscribed on the walls of our law colleges. It is a promise made by the judiciary to the people who come before it. The promise deserves a procedure worthy of its name.

Watch · The Office on YouTube

Two-minute explainers on the questions these essays raise.

The Office publishes short video versions of its long-form writing. Subscribe to be notified when the explainer for this article is released.

Sources and provisions referred to

Statutes and rules

Cases

Official documents

Frequently asked questions

What is judicial recusal?

The withdrawal of a judge from hearing a matter in which bias, personal interest, or the reasonable appearance of either would compromise the fairness of the proceeding. It rests on the maxim nemo judex in causa sua, no one should be a judge in his own cause, which together with the right to be heard forms the foundation of natural justice. Recusal protects not only the parties before the court but the institution itself, because justice must not only be done but be seen to be done.

Who decides whether a judge should recuse in India?

At present, the judge whose impartiality has been questioned decides the question personally, and no appeal lies against that decision as such. The essay identifies this as the paradox at the heart of the procedure: the very maxim that forbids judging one's own cause is suspended at the moment it matters most. Different judges have answered recusal requests in irreconcilable ways, some withdrawing readily, others treating the request itself as an affront, and the litigant has no way to predict which response awaits.

What reform does the essay propose?

A codified standard operating procedure with three elements: recusal requests made in writing on stated grounds, reasoned orders disposing of them, and reference of contested requests to a different bench, a Reference Bench, so that the question is not decided by the judge whose impartiality is in issue. The essay collects the scattered observations in the reported cases that already point in this direction and argues that consolidating them costs the institution nothing while protecting it from its most corrosive criticism.

What is the part-heard lock?

The practical rule that once a matter has been substantially heard by a bench, moving it elsewhere becomes almost impossible, because a fresh bench would have to begin again. The essay shows how this lock interacts perversely with recusal: a ground of bias discovered late arrives after the lock has closed, and the litigant must choose between waiving the objection and antagonising the bench that will decide his case. A codified procedure with early disclosure obligations would defuse most of these situations before the lock engages.

Can a litigant ask a judge to recuse, and is it risky?

A litigant can, and sometimes must, but the essay frames three duties that discipline the request: raise it at the earliest opportunity, raise it on specific grounds supported by material, and raise it respectfully, as an appeal to the institution's own standards rather than an attack on the individual judge. The risk is real, both of antagonising the bench and of being suspected of forum shopping, which is precisely why the essay argues the burden of deciding should be moved off the challenged judge's shoulders.

These answers are educational summaries of the article above; they are not legal advice on any individual matter.

Corrections: if you find an error of law or fact in this article, write to contact@lokeshbhardwajlawoffices.com with the subject 'Correction'. Corrections are made in the text and noted here with the date.

About the author and disclaimer

Lokesh Bhardwaj is an Advocate enrolled with the Bar Council of Delhi, practising before the Supreme Court of India and the High Court of Delhi.

This article is a contribution to the academic literature on natural justice and procedural reform of the Indian judicial system. It does not refer to, comment upon, or take any position with respect to any specific matter pending before any court, or any specific judge or bench. The principles, observations and proposals discussed are of general application to the institutional architecture of the Indian judiciary. The views expressed are personal and academic.

WhatsApp