A constitution had been adopted. Its authors had an electoral mandate, the required vote had taken place, and the country urgently needed a lasting settlement. On 6 September 1996, South Africa's Constitutional Court nevertheless declined to certify the text. The people who had written it would have to revise it.^[1]

This is an unusual place to pause in a story of political change. Constitution-making appears to be the point at which a community reaches the top of its own rules. Where could another binding question come from? And why should the authors have to answer it when they had been chosen through the country's first nonracial national election?

The election was real. So was the authority it conferred. What did not follow was that every provision written by the new majority was automatically sufficient. The rejected text makes the difference visible: having the standing to decide did not eliminate every condition on the decision.

Before anyone knew who would win

To understand the return of the draft, we need to go back to the transitional arrangement. Moving away from white minority control of national political power meant more than replacing the officeholders. It required agreement about who belonged in the electorate and how the powers of the new majority would be exercised.

The interim Constitution provided a bridge. Assented to in January 1994 and brought into force on 27 April, it established a common citizenship and nonracial franchise. The two houses of Parliament elected in 1994 would sit together as the Constitutional Assembly and write the final text. The negotiated interim rules and the subsequent constitution were distinct stages.^[2]

The sequence prevented either set of participants from simply standing in for everyone who came after. The negotiators could not settle every future question permanently through their own agreement. The elected representatives, in turn, did not acquire an unrestricted ability to discard the conditions under which the transition had been arranged.

Among those conditions were thirty-four Constitutional Principles. Section 71 of the interim Constitution required the final text to comply with them and to receive certification from the Constitutional Court before acquiring legal force. The judges' assignment was to test compliance with agreed principles, not to substitute the constitution they personally preferred.^[2]

That distinction did not make judicial judgment disappear. Principles need interpretation when applied to detailed provisions. But it gave the interpretation something it had to answer to. Which principle had not been met? What was missing from the text? A finding of noncompliance had to identify more than a judge's dissatisfaction with a political choice.

It would therefore be misleading to tell this episode as a court placing itself above the people. The court's authority was part of the transitional rules too. Nor did review erase the elected Assembly's authority to write. The arrangement made two forms of authority meet without allowing either one to do the entire job alone.

For someone newly admitted to an equal national franchise, that structure could matter beyond institutional design. A vote should change who gets to decide. The protection of someone who did not vote for the winner should not depend entirely on the winner's goodwill. Today's majority might itself become a minority on another question.

We should not pretend that this was what every voter in 1994 had in mind. An electoral record does not disclose private thought. What it does let us identify is a profound expansion in political standing, followed by a process that did not treat majority support as an answer to every subsequent question.

A refusal that required changes

The first certification proceedings heard objections from parties, organizations and individuals. The Court was not confined to the Assembly's account of its own work. In September it found that some provisions did not meet the prior principles, so the text as a whole could not yet be certified.^[1]

For a refusal to be more than another voice of authority, it has to become specific. Consider the Public Protector and the Auditor-General. Their functions included investigating government conduct and examining public money. The question was whether the draft sufficiently protected the independence of the people doing that checking.

The amended text strengthened appointment and removal requirements. A National Assembly recommendation for appointment would require at least sixty per cent support; a resolution calling for removal would require at least two-thirds, alongside the stipulated grounds and process. In December the Court found the revised safeguards sufficient for these offices.^[3]

Those fractions are not a universal recipe. Higher thresholds can also produce deadlock, and nomination and investigation procedures matter as well. The particular issue was more concrete: an office responsible for embarrassing questions should not retain security only for as long as those being questioned welcome the investigation.

The story does not belong to the Court alone. A refusal could have been ignored. A revised standard could have made compliance impossible. What happened instead was revision by the Assembly, resubmission, and certification on 4 December 1996. The main provisions of the final Constitution entered into force on 4 February 1997.^[3][4]

We learn something about a constraint when it actually makes a powerful decision-maker stop and alter the decision. Here the decision-maker had a mandate, a majority and compelling historical reasons for exercising authority. Those facts remained important. They did not settle the compliance question by themselves.

None of this establishes that courts are always right. Independent review is not a search for an infallible replacement ruler. It separates making a decision from being its only judge. A reader can value that separation and still question whether a particular judicial reason is persuasive.

The final Constitution also made room for questions initiated outside the main state institutions. Section 38 permits proceedings where a Bill of Rights guarantee is infringed or threatened, including certain claims brought for people unable to act for themselves, groups, the public interest and associations' members. The person directly harmed need not always carry the case alone.^[5]

Standing is not the same thing as affordable access. Time, legal help and the practical ability to pursue a remedy do not arrive automatically with a constitutional provision. We can recognize an enlarged entrance without pretending that everyone is equally able to reach it. Otherwise the legal text would once again be asked to answer for people's whole experience.

The families whose claim did not succeed

There is a tempting ending here: exclusion gives way to voting, an elected majority accepts limits, and rights gain routes to enforcement. But another case from the same year interrupts that smooth progression. Without it, the account risks becoming a diagram with the people edited out.

On 25 July 1996, the Court rejected a constitutional challenge to the amnesty arrangements. The applicants included AZAPO and members of the Biko, Mxenge and Ribeiro families. These were not anonymous costs attached to a successful transition. Their names appeared as parties before the court.^[6]

The issue extended beyond whether perpetrators would face prosecution. Amnesty could also remove civil liability for the relevant acts and affect claims against the state and other responsible bodies. A new political community was taking shape while particular routes through which families might pursue old injuries were being restricted.^[6]

Amnesty was not an automatic reward for invoking reconciliation. It required full disclosure of relevant facts and a qualifying association with a political objective. The Truth and Reconciliation Commission also had responsibilities concerning victims' accounts, the fate of those harmed and recommendations for reparation. The word amnesty should not swallow those conditions and procedures.^[7]

But discovering the truth and recommending reparation are not identical to preserving a claim for prosecution or damages. One may believe the exchange helped make a peaceful transition possible and still ask why some of its losses fell on people already injured. Accepting the first argument need not reduce the second question to an unfortunate footnote.

The Court held that the interim Constitution authorized the amnesty arrangement. It also acknowledged the unequal sacrifice between victims and perpetrators. That was a legal answer to the claim. It was not a personal conclusion that every affected family had to experience as reconciliation.^[6]

This is why it would be inaccurate to say that the new order expanded every route to accountability at once. Some people gained equal national political standing. Some could litigate in the public interest. At the same time, a transitional settlement limited particular claims arising from past wrongs. These changes cannot be averaged into a score without losing what happened to different people.

The opposite simplification would be to call the expansion fraudulent because the limitations existed. That, too, would make the history easier by deleting part of it. Exclusion was genuinely broken. Protections were genuinely added. Loss and unanswered demands remained genuine as well.

Nor did the Constitution's commencement certify that poverty, bereavement or inherited inequality had been resolved. A constitutional text can give later people grounds on which to act without performing those actions for them. Its continuing use belongs to lives that cannot be completed by the date on which it took effect.

Two documents are worth holding together: the constitution returned for revision and the families' unsuccessful application. One shows a new majority accepting a constraint on its rule-making. The other keeps the settlement from becoming a gift for which no one had to pay an additional price.

Before deciding whether a state has advanced or retreated, we can ask smaller questions. Who can now participate? Who can require an answer? Which challenges actually change a decision, and which claims are restricted in the name of a shared future? The questions do not finish the judgment for us. They make it harder for our judgment to pass over the people inside the history.