The Reference That Should Not Have Been Made: Bela M. Trivedi's Dissent on Sub-Classification and Stare Decisis

On 1 August 2024 a seven-judge bench overruled <em>E.V. Chinnaiah</em> by 6:1 and permitted States to sub-classify within Scheduled Castes. The lone dissenter, Bela M. Trivedi, devoted the first quarter of her opinion not to Article 341 but to whether the reference making the overruling possible should ever have been entertained. A reading of her three-part dissent.

On 1 August 2024, a seven-judge bench of the Supreme Court overruled a five-judge Constitution Bench decision by a vote of six to one. The Constitution Bench was E.V. Chinnaiah v. State of Andhra Pradesh (2005). It had held that the conglomeration of castes in the Presidential List under Article 341 of the Constitution forms a single homogeneous class, that States have no legislative competence to sub-classify it for reservation purposes, and that any such sub-classification violates Articles 14 and 341 read together. Chinnaiah stood for fifteen years. In State of Punjab v. Davinder Singh, the seven-judge bench overruled it. The single dissenter was Justice Bela M. Trivedi.

Her dissent is in three parts, and the first part is not about Article 341 at all. Across paragraphs 17 to 25, before reaching any substantive question on the Presidential List or the meaning of “Scheduled Castes,” Trivedi develops a procedural argument: that the reference itself — by a three-judge bench in 2014, sending a Constitution Bench’s decision to a larger bench for reconsideration — was improperly made. The order making the reference was, in her words, “very cryptic and perfunctory” and “not supported by any reason.” The Five-Judge Bench that received it in 2020 should not have countenanced it. The Seven-Judge Bench that overruled Chinnaiah on its strength was acting on a reference that the doctrines of precedent and stare decisis would not have permitted.

The line in which the position crystallises is at paragraph 25: When a law was settled by the previous Constitution Bench in E.V. Chinnaiah after considering all the previous judgments including Indra Sawhney, and after investing substantial judicial time and resources, and when the same had held the field for a substantially long period of fifteen years, in my opinion, the very reference by the Three-Judge Bench to the larger bench for reconsideration of the decision in E.V. Chinnaiah, that too without assigning any reason was inappropriate and not in consonance with the well settled doctrines of Precedents and Stare decisis.¹

A dissent that begins with bench discipline rather than with the constitutional text on which the case nominally turns is a particular kind of dissent. Its first claim is not that the majority has read the Constitution wrongly. Its first claim is that the majority should not have been asked the question.

The doctrine the dissent invokes

The argument moves through five authorities, each cited for a slightly different proposition about when settled decisions of larger benches may be reopened.

The first is Pradip Chandra Parija v. Pramod Chandra Patnaik, a 2001 Five-Judge Bench decision, which set out the procedural floor: a smaller bench that doubts a larger bench’s decision must, at minimum, set out reasons for disagreement and refer the matter to a bench of the same strength as the one that decided originally — not directly to a larger bench. The 2014 three-judge order in Davinder Singh, on Trivedi’s reading, did neither. It set out no reasons, and it referred a five-judge bench’s decision directly to a constitution bench of larger strength.

The second is Maganlal Chhaganlal (P) Ltd. v. Municipal Corporation of Greater Bombay, a 1974 Seven-Judge Bench decision in which H.R. Khanna J. wrote that “certainty in the law, which was an essential ingredient of the Rule of Law, would be considerably eroded if the highest Court of the land lightly overruled the view expressed by it in earlier cases.” Khanna J. allowed one exception: overruling could be appropriate “where contextual values giving birth to the earlier view had subsequently altered substantially.” Nothing in the Davinder Singh reference suggested that the contextual values underlying Chinnaiah had so altered.

The third is Lt. Col. Khajoor Singh v. Union of India, another Seven-Judge Bench, which laid down what amounts to a unanimity threshold: the Court should not depart from an earlier interpretation unless there was “a fair amount of unanimity that the earlier decision was manifestly wrong.” The fourth, Keshav Mills Co. Ltd. v. Commissioner of Income Tax, also a Seven-Judge Bench, added that frequent review of earlier decisions “on the ground that the view placed before it later appeared to the Court to be more reasonable” makes the law uncertain. Both decisions place revisitation beyond the threshold of mere reasoned disagreement.

The fifth is Dr. Shah Faesal v. Union of India, a 2020 Five-Judge Bench decision that reiterated the doctrines of precedent and stare decisis as a continuing feature of constitutional adjudication, drawing the line that Trivedi adopts: a bench of smaller strength must, before doubting a larger bench’s decision, record reasons that meet a substantive threshold of justification, not merely a feeling that the earlier reading is less attractive.²

Taken together, the five authorities trace a continuous line. Each places revisitation behind a threshold of justification. None permits the kind of casual reference Trivedi describes the 2014 order as making. The chain stretches from H.R. Khanna J. in 1974 to a five-judge bench in 2020. It was, by 2024, doctrinally settled.

What the 2014 order said

The order Trivedi describes as casual and cavalier ran to a few hundred words. It was passed by a three-judge bench on 20 August 2014. The relevant passages, reproduced in her dissent, observe that the Additional Solicitor General submitted that Chinnaiah “has no application on the controversy in hand” and was “not in accord with the 9-Judge Bench decision of this Court in Indra Sawhney.” The bench then directed that the matter be placed before a larger bench. It did not adopt the Additional Solicitor General’s contention, expressly. It did not engage with the reasoning of Chinnaiah. It did not analyse the relationship between Indra Sawhney (a 1992 nine-judge bench on the Other Backward Classes question) and the Scheduled Castes carve-out that Chinnaiah had insisted on. It referred.

The objection Trivedi raises is procedural in character but doctrinal in stakes. If a Constitution Bench’s settled decision can be sent to reconsideration on a few hundred words of order text that does not engage the case’s reasoning, then the threshold for revisitation has dropped from the substantive standard the seven-judge authorities had set down to a permissive standard requiring little more than a request and a referral. The certainty in the law that H.R. Khanna J. had warned against eroding in 1974 had been eroded by the time the reference was made.

Whether this is the right reading of the 2014 order is itself a doctrinal question. Defenders of the majority would say that the three-judge bench did identify a substantive ground for revisitation — the possible inconsistency between Chinnaiah and Indra Sawhney — even if it did not develop that ground at length. The dissent’s position is that identification without development is exactly the casual referral the seven-judge authorities had prohibited. The two readings of the 2014 order disagree about how much reasoning the procedural threshold requires.

Article 341, and the textualist case for Chinnaiah

The dissent’s second part — paragraphs 26 to 57 — turns to the substantive question. If the reference is entertained on its own terms, was Chinnaiah correctly decided?

Trivedi’s answer is textualist. Article 341 of the Constitution has two clauses. Clause (1) gives the President power, after consultation with the Governor, to specify by public notification “the castes, races or tribes or parts of or groups within castes, races or tribes” which shall, for the purposes of the Constitution, be deemed to be Scheduled Castes in relation to a State or Union Territory. Clause (2) provides that Parliament may, by law, include in or exclude from the list specified in the notification any caste, race, or tribe. The clause then closes with what Trivedi treats as the operative restriction: save as aforesaid a notification issued under the said clause shall not be varied by any subsequent notification.

The restriction is absolute on its face. Variation of the Presidential List is reserved to Parliament alone, exercised by law. Not by the President himself through subsequent notification. Not by State Legislatures. Not by State executives. Not by the Court.

The constitutional debate the dissent reproduces from the Constituent Assembly proceedings of 17 September 1949 supports this reading. Dr. Ambedkar, introducing what became Article 341, explained that the President was to issue a notification specifying the Scheduled Castes; “the only limitation put was that once a notification has been issued by the President, any elimination from or any addition in the list must be made by the Parliament and not by the President.” The purpose, Dr. Ambedkar continued, was “to eliminate any kind of political factors having a play in the matter of the disturbance in the Schedule so published by the President.” The Presidential List, in other words, was meant to be insulated from local political contestation — both from the President’s own subsequent variation and from State-level reorganisation.

Once Article 341 is read this way, the substantive case against State sub-classification follows. A State law that divides the Presidential List into groups and apportions reservation quotas among them does not, on its face, add or remove a caste. But it varies the constitutional status the Presidential List confers. A caste that the President has placed within the Scheduled Castes for the purposes of the Constitution acquires, by that act, a single legal character. To divide that character into preferred and non-preferred sub-classes — even with the purpose of advancing the most backward — is, on the textualist reading, to vary the notification.

The doctrinal route by which the majority allowed sub-classification — reading State competence under Articles 15(4) and 16(4), or under enabling provisions in the State List, as extending to rationalisation of reservation within the Scheduled Caste class — runs, for Trivedi, into the colourable-legislation bar. “Under the guise of providing reservation or under the pretext of taking affirmative action for the weaker of the weakest sections of the society, the State cannot vary the Presidential List and tinker with Article 341.” The doctrine she invokes is one of the older ones in constitutional adjudication: you cannot do indirectly what you cannot do directly.³

Why the reasoning matters beyond Davinder Singh

Article 341 is a narrow provision. Sub-classification within the Scheduled Castes is a specific question. But the textualist case the dissent develops rests on a wider proposition: that where the Constitution reserves a power to one institution — in this case, to Parliament under Article 341(2) — other institutions cannot exercise that power, however benevolent the purpose, through the doctrinal devices that purposive interpretation makes available.

The proposition runs through several lines of constitutional adjudication. It runs through the cases on Article 105 and parliamentary privilege, where the Court has declined to read into general phrases reservations the Constitution has placed elsewhere. It runs through the cases on Articles 32 and 226, where the Court has been reluctant to permit one writ jurisdiction to do work the Constitution has assigned to another. The Article 341 case is one application of a broader principle about constitutional allocation. The dissent is doing more, in this respect, than defending Chinnaiah. It is defending a method of constitutional reading against an alternative method.

The alternative method is the purposive one the majority adopts. On that method, the question is whether sub-classification advances the substantive aim of Articles 14 and 16 — the equitable distribution of reservation among the most backward. The textualist objection is that the substantive aim does not authorise institutional rearrangement. The Constitution’s drafters placed the variation power in Parliament. The Court cannot relocate that power, however well it might be exercised elsewhere, without violating the allocation itself.

Indra Sawhney and what it did not decide

The dissent’s third movement — paragraphs 58 to 76 — takes up the question on which the 2014 reference rested: whether E.V. Chinnaiah’s prohibition on sub-classification of the Scheduled Castes is consistent with the reasoning of Indra Sawhney v. Union of India, the 1992 nine-judge bench on Other Backward Class reservation.

Trivedi’s position is that the two cases addressed different questions and do not pull in opposite directions. Indra Sawhney dealt with the meaning of “backward class of citizens” in Article 16(4) and the constitutionality of the 27% reservation for Socially and Educationally Backward Classes under the August 1990 office memorandum. The Court did permit, in that context, sub-classification within OBCs and the creamy-layer exclusion. But its reasoning was explicitly limited. Jeevan Reddy J., writing for four of the nine judges, observed that “whenever such a situation arises in respect of Scheduled Castes, it will be Parliament alone to take the necessary legislative steps in terms of Clause (2) of Article 341.” The States, in Indra Sawhney’s own framing, did not have the legislative competence. The carve-out for OBCs was a carve-out, not a general principle.⁴

The dissent works through the concurring opinions in Indra Sawhney — Pandian J., Sawant J., and the others — to show that none of them treated their OBC reasoning as transposable to the Scheduled Castes. The conclusion at paragraph 75 is unambiguous: “in none of the cases — Indra Sawhney or Jarnail Singh, the issue of sub-classification of ‘Scheduled Castes’ in the context of Article 341 was raised or argued, nor was decided by the concerned Benches.” The 2014 order’s premise — that Chinnaiah was “not in accord” with Indra Sawhney — was, on the dissent’s reading, a misreading of what Indra Sawhney had said.

This part of the dissent is the part that returns, in a sense, to the procedural argument with which it began. If Indra Sawhney did not decide the Scheduled Castes question, the doctrinal inconsistency that the 2014 reference identified did not exist. The reference rested on a premise that closer reading of the case did not support. The casualness of the reference, on this reading, was not only procedural but substantive: it failed to engage with what Indra Sawhney had actually held.

Article 142, and the limit of equitable jurisdiction

The dissent’s closing movement is short, and it is its sharpest. Paragraphs 77 and 78 take up the question of whether the majority’s decision can be justified by reference to Article 142 of the Constitution — the provision empowering the Supreme Court to pass any order necessary for doing “complete justice” in any cause or matter pending before it.

Trivedi reaches back to Supreme Court Bar Association v. Union of India, a 1998 Constitution Bench decision on the limits of Article 142. The SCBA judgment had held, in the context of contempt proceedings, that Article 142 “cannot be used to build a new edifice where none existed earlier, by ignoring express statutory provisions dealing with the subject and thereby to achieve something indirectly which cannot be achieved directly.” The phrase echoes the one Trivedi has already used in the colourable-legislation passage. The two operate as bookends on the same argument: institutional power, whether legislative or judicial, cannot be exercised to circumvent the Constitution’s specific allocations.⁵

The application to Davinder Singh is direct. If States have no legislative competence to vary the Presidential List under Article 341, the Court cannot, under Article 142 or any related equitable jurisdiction, validate State action that would otherwise be ultra vires. The width of Article 142 is amplitude, not authority to dispense with constitutional text.

The action of the State though well-intentioned and affirmative in nature, if violates the specific provision of the Constitution, cannot be validated by the Supreme Court in exercise of its jurisdiction under Article 142. The removal of inequalities or remedy to remove inequalities cannot be permitted at the cost of violation of the specific provision of the Constitution.Bela M. Trivedi J., para 78 of the dissent

The line condenses the dissent’s philosophical commitment into a single proposition. Affirmative intent does not, in the Trivedi reading, license constitutional rearrangement. The Court’s role is to enforce the allocation the Constitution has made, including where that allocation places certain powers beyond the reach of the institutions that might be tempted to use them most generously. This is a position about judicial method; it is also a position about institutional humility.

Reading the three layers together

The dissent’s three movements — the procedural argument, the Article 341 textualism, the Article 142 limitation — are not separate arguments. They are three aspects of a single position about how constitutional courts should treat their own jurisdiction.

The procedural argument concerns horizontal discipline: how the Court treats its own past decisions across benches of different strengths. The Article 341 argument concerns vertical discipline: how the Court reads the constitutional allocation of powers between Parliament, the President, the States, and itself. The Article 142 argument concerns the boundary between adjudication and legislation: how far the Court’s equitable jurisdiction can travel when constitutional text closes a route.

What unites them is a common premise about the limits of judicial authority. Trivedi’s position is not anti-affirmative-action. She does not argue that Scheduled Caste reservations are wrong, or that the most backward among the backward should not receive specific protection. Her position is that the institutional route to that protection is Parliamentary legislation under Article 341(2), not State sub-classification ratified by judicial overruling. Where the Court has overruled Chinnaiah, it has, on her reading, taken on a legislative role the Constitution reserved to another body.

The opening of her dissent, in paragraphs 1 to 4, frames the position in unusual terms. She quotes Justice William O. Douglas of the United States Supreme Court — “a great dissenter who had written as many as 486 dissenting opinions” — on the right to dissent as that which “gives dignity, worth, and individuality to man.” She then quotes Justice Oliver Wendell Holmes, dissenting for the first time in Northern Securities Company v. The United States in 1903: “I am unable to agree with the judgment of the majority of the Court,” Holmes wrote, before pressing on to explain why he felt bound to record his reasons nonetheless. The opening is more than rhetorical. It marks the dissent as one of those that records its reasons not in the expectation of reversing the result but in the expectation of preserving an alternative reading for later use.

What the dissent preserves

The seven-judge bench’s decision is the law. As of August 2024, States may sub-classify within Scheduled Castes for reservation purposes, subject to such constraints as the majority’s reasoning imposes. The administrative consequences are working themselves out in State legislative practice; Punjab, Tamil Nadu, and other States that had attempted sub-classification before Chinnaiah or after it have begun to revisit their statutes.

What Trivedi’s dissent preserves is twofold. The first is doctrinal: an articulated textualist reading of Article 341 that future benches addressing variation of the Presidential List — whether on Scheduled Tribes under Article 342, or on Socially and Educationally Backward Classes under Article 342A, or on questions yet to arise — may draw on. The reading does not require Chinnaiah to remain good law in order to be doctrinally available. It is anchored in constitutional text, in Constituent Assembly material, and in a chain of decisions on the exhaustiveness of the Presidential Lists.

The second is meta-judicial: a statement of the limits within which a Constitution Bench should treat its own settled decisions as open to revisitation. The chain of authorities Trivedi assembles — Maganlal Chhaganlal, Khajoor Singh, Keshav Mills, Pradip Chandra Parija, Shah Faesal — was not, in 2024, a contested chain. The dissent treats it as binding. Whether subsequent practice treats it as binding remains to be seen. The Davinder Singh majority itself does not engage with the chain at length.

One observation closes the dissent’s force. Of the three Constitution Bench decisions overruled by the Chandrachud court in 2024 — India Cement in Mineral Area Development Authority, Synthetics and Chemicals in Lalta Prasad Vaish, and now Chinnaiah in Davinder Singh — each had stood for between fifteen and thirty-four years. Each had been operated by States and central administration as settled law. Each had been the subject of administrative architecture, of statutory amendment, or both. The 2024 court, in one calendar year, revised all three. The threshold the seven-judge authorities had set down required “a fair amount of unanimity that the earlier decision was manifestly wrong.” The dissent in Davinder Singh records, for any future bench that may examine the question, what unanimity at that threshold would have looked like, and what its absence cost.