Sabarimala reference | Who on the Bench engaged, and when

In May 2026, a nine-judge bench led by Chief Justice Surya Kant reserved judgement in the Sabarimala Reference after 16 days of argument. The reference came from a review petition challenging the correctness of the 2018 decision that had lifted the ban on women of menstruating age entering the Ayyappa temple. We previously broke down the time taken by counsel for submitting their arguments. In this data-piece, we look at the other side: who on the Bench engaged, when, and for how long.

The Bench spoke 91,297 words across the 16 days. This amounted to 16.2 per cent of the hearing’s 562,722-word total. That share is a little higher than the nine-judge bench—13 per cent—recorded in the definition of “industry” reference heard weeks earlier. 

The distribution in the Sabarimala Reference was steeply uneven: four out of nine judges engaged the most while three barely spoke at all.

Figure 1: Words and turns for each judge, all 16 days. Bars sorted by words; the three near-silent judges sit at the foot.

The active Bench was four judges

Figure 1 sets each judge’s volume of speech against the frequency of intervention. The two are not the same measure, and the gap between them is where the reading lies. 

Four judges — Justice B.V. Nagarathna, Justice M.M. Sundresh, CJI Surya Kant and Justice Joymalya Bagchi — together accounted for 85 per cent of the Bench’s words. Add Justice Ahsanuddin Amanullah, and the five judges supplied 97 per cent.

The most striking figure is at the top. Justice Nagarathna, not the presiding CJI, was the most active judge on the Bench by both measures: 22,527 words across 1,868 turns, a quarter of all the Bench’s words and nearly a third of its turns. CJI Surya Kant came third by volume, at 19,475 words. A presiding judge speaking less than a puisne colleague is not the usual pattern; in the industry reference the CJI had led his Bench comfortably. Here the floor was managed differently, and the record shows it plainly.  

Volume and frequency diverge sharply among these four. The Chief Justice and Justice Nagarathna intervened often and briefly, at 11.5 and 12.1 words per turn: short interjections, floor management, calling on counsel, a sentence of doubt. Justice Sundresh and Justice Amanullah did the opposite, at 20.4 and 22.7 words per turn: fewer interventions, each carrying a developed question. Justice Bagchi spoke 14,222 words across 738 turns—19.2 words per turn. Justice Sundresh took the second-highest word count, 21,041, in barely more than half Justice Nagarathna’s turns. When he spoke, he spoke at length. 

The shape of the 16 days

Figure 2 plots the Bench’s share of each day’s words.It fell to 8.6 per cent on 17 April, the quietest day. It rose to its peak of 26.9 per cent on 7 May, when the parties opposing the review presented their case and the judges tested submissions hardest.

Rejoinders began on 13 May. That day saw the single heaviest raw volume—37,356 words—yet the Bench’s share fell to 17.3 per cent: as counsel pressed their closing arguments, the judges largely listened Engagement, in other words, tracked not the length of a day but the point in the argument at which a proposition needed testing.

Figure 2: The Bench’s share of each day’s spoken words. 

The busier days cluster in May, when the rejoinder and the final submissions drew the Bench in. The single heaviest day of argument by raw volume was 13 May, at 37,356 words, yet the Bench’s share that day was 17.3 per cent: counsel spoke a great deal, and the judges let them. Engagement, in other words, tracked not the length of a day but the point in the argument at which a proposition needed testing.

On the three who barely spoke

At the top of Figure 1 sit three judges who were, in practice, silent. Justice P.B. Varale spoke 980 words across the whole reference; Justice R. Mahadevan, 283; and Justice A.G. Masih, just 19 words in two interventions across 16 sitting days. Together the three spoke 1,282 words, under a quarter of one per cent of the hearing. A fourth, Justice Aravind Kumar, spoke more — 1,636 words — but in short bursts, and rarely carried an exchange. 

Silence on the dais does not measure a judge’s contribution. There is no rule in the Indian Supreme Court that a judge must speak for their participation to count, and the eventual judgement will be authored, joined or dissented from by all nine on equal terms. In the commentary, the citations and the reasoning that follow this reference, the voices that will recur are the four who carried it. The others will appear at the head of the reportable judgement, and may yet write opinions of their own.

“Without hearing millions”

The pattern of long, testing questions from a few judges shaped the reference’s central anxiety: by what standard may a court judge a religious practice at all. On the fourth day, as Senior Advocate Abhishek Manu Singhvi pressed the case against the essential religious practices test, Justice Sundresh put the difficulty in a single question — whether a court could rule on the validity of a religious practice, in his words, “without hearing millions” who follow it. The Chief Justice framed the same worry, observing that the “most difficult task” for a constitutional court is to declare that a belief held by large numbers is wrong.

The exchange is characteristic of how this Bench worked. The questions that gave the reference its shape came not from the Bench as a body of nine but from four of its members, returning again and again to the same fault lines: the reach of judicial review into faith, the meaning of constitutional morality, and the line between an individual’s right under Article 25 and a community’s under Article 26. When the judgement comes, it will speak for all nine.

Methodology

The figures are drawn from the 16 TERES argument transcripts published by the Supreme Court for the hearings dated 7, 8, 9, 15, 17, 21, 22, 23, 28 and 29 April and 5, 6, 7, 12, 13 and 14 May 2026. Each speaker’s distinct stretches of speech were extracted, and the words counted after removing speaker labels and the transcriber’s bracketed annotations such as [UNCLEAR]. A ‘turn’ is the speech of one named speaker between two speaker changes. Estimated times, where given, use a conventional 140 words per minute; the transcripts carry no timestamps, so times are estimates, not measurements. Word counts are robust to within a small margin that depends on how labels and page furniture are stripped, and are reported as such rather than to the exact word.

Data extraction and counting were performed by a script written for this analysis; its output was cross-checked on a second independent pass, and the per-day totals reconciled against the grand totals. All editorial judgements and the conclusions drawn from the data are the author’s.

N.V. Geetha ran the data extraction on which this analysis is based.