Ghost Voices

Last month's Feature was about audiences who fought over a play. This one is about audiences who did not know there was anything to fight about — until somebody told them.

Hollywood calls her a ghost. India calls her playback, and means it as a job description. The technique is as old as the sound film, and in India, by the best accounts we could find, it began as a repair job — a chorus too tired to sing and dance at once. What these six pieces are about is the part that isn't technical: who was told, who was credited, and who was paid.

Look closely and the complaint is never about the voice. The Milli Vanilli songs were hits — three number ones. The voice in Beijing was, in its own music director's word, perfect. Nixon's was good enough that three of Hollywood's biggest musicals were built on it. The trouble arrived with the information.

Six pieces, in the order they have to be told. A song recorded before it was filmed, in Calcutta in 1935. A record in 1949 with the wrong name on it. A soprano who was the unseen voice of Hollywood's leading ladies through the 1950s and '60s. A pop duo whose award was taken back after the voices on its record turned out not to be its own. A Politburo member at a rehearsal in Beijing, and a decision about which child the nation would hear and which it would see. And a fight over a royalty that divided Lata Mangeshkar and Mohammed Rafi in the 1960s, long before Parliament strengthened performers' rights in 2012 and singers formed a society to collect what they were owed.

A ghost is not a voice without a body. It is a voice without a name.

Recorded first. Filmed second. Voice, then face — that order is the whole history of playback in a sentence, and it was set down in Calcutta, at New Theatres, for a Hindi film called Dhoop Chhaon. The film was a remake of a Bengali one, Bhagya Chakra, released the same year. Nitin Bose directed. R.C. Boral composed — Pankaj Mullick is named beside him in some accounts — and Boral is the man Anil Biswas would later call the "Bhishma Pitamah" of film music. Most histories give him the credit for bringing playback to Hindi cinema.

Before that, songs in Indian films were sung live, on camera, by the people on screen. You sang; the camera watched you sing. Playback, as one writer on the period puts it, "freed the artist from the bondage of the microphone."

How it came about is wonderfully unglamorous. A group dance of sakhis — a chorus of women, companions to the heroine — was being rehearsed for the two films, and the rehearsals took so much out of the performers that when the cameras ran, their singing went haywire. Singing and dancing at once looks easy from the stalls. So the fix, as the film-song historians tell it, was Bose's: record the song first, play it back on the set, and let the chorus dance to it. Boral made it work with Mukul Bose, Nitin's brother, the sound recordist whom the Indian Cinema Heritage Foundation's Cinemaazi archive credits with helping execute the playback experiment. The director had an idea. Somebody at a recording desk had to make it possible.

Playback, at its birth, was a way of keeping a voice steady — not a way of replacing it.

Who sang? Parul Ghosh, Suprabha Sarkar, Harimati Dua — and, in the Hindi version, a blind singer called K.C. Dey, uncle of the later playback star Manna Dey. Sarkar was, by one account, a Shantiniketan student at the time. Ghosh was born in Barisal in 1915; her brother was the composer Anil Biswas, who introduced her to playback, and she would sing in Hindi and Bengali films until 1951.

Ask who the first playback singer in Hindi cinema was, and the answer depends on whom you read. One account makes the first three — Ghosh, Sarkar and Dua — the pioneers. Another says Sarkar and Dey were the first to sing for somebody else, while Ghosh and Dua sang for themselves.

The first playback singers in Hindi cinema may have been singing for themselves.

If all the singers were supplying their own screen voices, this would be playback without a ghost. If some were singing for others, the separation was there already. Recording before filming and replacing the person who sings are two different things, even when the histories give them the same name.

Every "first" in this story is like that. Bombay has its own claimant: Mahageet (1937), Sagar Movietone, a song sung, by one account, by the composer Anil Biswas himself. Another goes back further, to a 1931 film called Farebi Jaal. A record collector writing about the 78 rpm era notes that a singer named Master Vasant Amrut of Surat is credited with that film's lyrics, sang three of its six songs, and is also listed in the cast — which probably makes him an actor who sang, not a playback singer. And the collector says plainly why he can't be sure: the reels of Farebi Jaal are not available, so nobody can check whether the songs on the discs were the songs in the film. (His source is Harmindar Singh Hamraaz's Hindi Film Geet Kosh, a song-by-song reference compiled in 1988 for exactly this reason.) Somebody remembers a song. Somebody else remembers a record. In this case, the film is gone.

The studio is part of the joke. New Theatres in the 1930s was a studio of singing stars: Kanan Devi was its number-one singer-star, on a par with K.L. Saigal, who was known as a singer and an actor in equal measure. The studio that gave Indian cinema the voice that isn't in the picture was a studio whose stars were their own voices.

For some years after Dhoop Chhaon, that is how it stayed. Between 1935 and 1940, by one reckoning, it was mainly actors who recorded their songs first and mimed to the playback. Trained singers came to the form slowly. Rajkumari and Amirbai Karnataki, actor-singers, either gave up acting or cut back to concentrate on playback; with Zohrabai Ambalewali and Shamshad Begum, they are counted among the first generation of the Hindi film's professional playback singers. What turned the ghost into an industry was not art. It was the shop counter. Producers noticed that gramophone records of songs sung by trained professionals sold like hot cakes — and went looking for the professionals. They looked first at All India Radio. (The same radio that August's Feature, What the Air Doesn't Keep, mourned for the things it failed to hold on to was where the film industry went shopping for voices. Most of the popular playback singers of the 1940s began there.)

A voice that isn't the face's is a technique. It becomes a ghost only when the audience isn't told.

And on that, the record goes quiet. The people who now do the work of identifying 1930s and 1940s film singers — collectors, reference books like the Geet Kosh — are still at it; songs of that period turn up with the singer, the lyricist, the composer unidentified. One blogger, trying to work out who sang a song from the 1948 film Chupke Chupke, found that the label of his 78 rpm record did not mention the singer at all. Some websites say Iqbal Bano. He points out that she was born around 1935. So he wrote to an expert on the period's singers, Girdharilal Vishwakarma, and for the time being the song is credited to a singer called Nazira Begum, who recorded songs for only four films, three of them unreleased.

Fourteen years after a tired chorus in Calcutta was recorded first and filmed second, a record would go on sale in India with a name on its label, by the singer's own account.

It was not hers.

Kamini was not a singer. Kamini was the heroine of Mahal (1949) — played on screen by Madhubala, directed by Kamal Amrohi, scored by Khemchand Prakash. The singer was twenty years old. Her name was Lata Mangeshkar. The disc carried the name of the character the song belonged to. The voice had been filed under the face.

It is a wonderfully neat ghost story. The evidence is less neat, but there is more of it than a single recollection.

Lata tells her part in her conversations with the film historian Nasreen Munni Kabir. She remembers the Kamini label. She remembers asking producers to credit the singers. And she remembers listeners writing to the radio to ask who was singing a song they could not get out of their heads. The audience had a voice it recognised and a name it did not.

(This is the thread August's Feature, What the Air Doesn't Keep, kept pulling on: radio as the place where India's voices were heard and, mostly, not kept. Here the listeners wanted the name to travel with the sound.)

Then there are the collectors. The 2003 issue of The Record News, preserved by the Digital South Asia Library, describes two versions of the 78 rpm record: one naming Kamini, the other Lata Mangeshkar. Its account says she insisted on having her name printed. A collector's description is not the same thing as examining both discs ourselves. It is, however, evidence about the objects, and it belongs beside the singer's memory.

Two labels. One voice. A character's name and a woman's.

There are three questions here, and they should not be made to do each other's work. Was the record issued under Kamini's name? Lata's testimony and the collectors' account support that. Was it also issued under Lata's? The collectors describe that version too. Exactly when did the change happen, and which request finally made the company act? Those are the questions these accounts do not date precisely enough to close.

The letters matter. So does the singer asking to be named. It is tempting to make one listener's envelope the hinge on which the whole industry turned. But Lata's account is also an account of work: asking producers, objecting, asking again. A public wanted an answer. A singer wanted a credit. Those are related demands, and they need not have arrived at the same desk on the same morning.

Notice what had been put on sale. Fourteen years after a tired chorus in Calcutta was recorded first and filmed second (piece 01), the object going into people's homes could still carry the heroine's name where the singer's should have been. The film's makers knew who sang. The purchaser was given Kamini. Kamini was a woman who did not exist.

Which is why the ghost in Hollywood and the ghost in Bombay are not quite the same ghost. Marni Nixon (piece 03) describes secrecy enforced by a studio. Lata describes a system in which singers' names were left out of the credit, and a fight to put them in. We do not need to imagine what every person at the record company intended. The label tells us what the purchaser was told.

In the same conversation with Kabir, Lata identifies Barsaat, Andaaz and Badi Bahen, all from 1949, with the arrival of named credits on screen and disc. That is her account of the change, rather than a chronology established here by inspecting every original release. A streaming service's tidy credits cannot do that work retrospectively. They show what is said now. The old discs show what was said then.

So the story has a second half, and it is the half where the name wins. Her name goes on the record. Within a few years there is hardly a film that does not want it.

But think about what had to happen first. A singer had to ask for something that now looks too obvious to require asking. Listeners had to write for information a label could have supplied in a line. The voice was already doing its job. The credit had to catch up.

Seventy-seven years on, her name on "Aayega Aanewala" is not in doubt. The collectors' account preserves the fact that another name was there too. What remains less clear is the date on which one gave way to the other.

That is the thing about a name. It can be wrong on one pressing and right on the next. The song sounds exactly the same.

She stood in a recording studio with an orchestra and sang the songs of West Side Story — all of them, the whole part — believing, as Marni Nixon remembers it, that her own track would be blended with Nixon's for the high notes. Nixon was in the same studio, singing the same songs. Wood was praised, session after session. Then the film's makers threw her track away.

"It ended up being totally my voice," Nixon said, years later, to the critic Jasper Rees. The line has no bitterness in it. It is the line of a professional reporting a result.

Nixon's co-author, Stephen Cole, is blunter about what happened to the actress. She was lied to, he says, from day one. Wood had been told one thing, and the studio had decided another, and the person who knew both sat in the next booth, a soprano from California, having the unusual experience of being flattered at one side of the glass and erased on the other.

That is the Hollywood version of the ghost. It is worth sitting with how different it is from the one in the last piece.

Lata Mangeshkar was left off a record label and had to fight to be named. Marni Nixon describes a studio that required her to keep quiet. Both arrangements separated a voice from its owner in public. In Hollywood, the silence was part of the instructions.

Start with the threat. After she sang for Deborah Kerr in The King and I (1956), Nixon says, 20th Century Fox warned her that "they would see to it that I wouldn't work in town again" if she told. She was frightened, she says, and she kept quiet. She had, according to Cole, signed a contract that bound her to silence. Kerr, he says, told her years later that she herself had never signed anything of the kind. The star didn't know the rules her ghost was living under.

When people asked, Nixon had a line for it. She had to "take the Fifth Amendment."

It is an odd and rather chilling joke about keeping quiet. She was not before a committee or in a courtroom. She was a singer being asked whose voice an audience had heard, and her livelihood depended on not answering.

So the secret was kept by contract and by fear, and it was kept for a good reason. The studios were selling more than a film. They were selling a woman, with a face and a voice together, and the marketing department had no wish to explain that the two had been bought separately. A star was a package. The voice that completed it could be paid for without being publicly named.

Which brings us to the second part of the story: what the ghost was paid. For West Side Story, the official production history describes Nixon as employed day by day, without a signed contract. When she sought a share of the soundtrack royalties, the film and record producers refused. Leonard Bernstein broke the impasse by giving her a fraction of his own royalty. Nixon recalled the arrangement herself in a 2015 interview.

She did receive royalties. The distinction is how. The people selling the soundtrack would not give her a share; a composer gave up part of his. It was a generous act, and the need for generosity is the problem. What would have happened to a singer without a Bernstein? (Piece 06 returns to that question in Bombay, where Lata Mangeshkar wanted a royalty that did not depend on a favour.)

The work itself had begun long before Maria. In Joan of Arc (1948), she was one of the angel voices heard by Ingrid Bergman's character — not Bergman's singing double. She sang for Margaret O'Brien in The Secret Garden (1949), and supplied high notes for Marilyn Monroe's "Diamonds Are a Girl's Best Friend" (1953). These were different jobs, gathered in the Masterworks Broadway account of her career under a familiar word: dubbing.

In West Side Story she sang Maria and, at one recording session, Anita too. In her account to the film distributor Park Circus, both Rita Moreno and Betty Wand, who also dubbed Anita, were ill when the quintet was being recorded. Nixon supplied both Maria and Anita, changing the sound of her voice. A duet with herself inside an ensemble. That is as much a piece of craft as any film acting of the period. It is also the point at which the arithmetic gets absurd. Two characters. Two famous actresses. A voice the film did not name.

The woman with the most to lose in this story, though, is not Wood. It is Audrey Hepburn.

For My Fair Lady (1964) Hepburn took singing lessons and went back into the studio again and again. She expected, according to Cole, that most of her own vocals would be used. She kept trying. Nixon says Hepburn could tell she wasn't making it. Of the three leading ladies Nixon sang for, she says her working relationship with Hepburn was the better one.

The thing to notice is that Hepburn seems to have known. Wood, by Cole's account, had been lied to. Kerr, by his account, hadn't been told the rules. Hepburn spent months trying to sing Eliza, and she was the one who had to feel the voice slip away in the room.

Then the secret broke. It didn't break in a scandal. It broke the way these things do, in a murmur. After West Side Story and My Fair Lady, Nixon says, "it got to be well known that this sort of thing was going on in Hollywood." Audiences had been asking why these women sounded so good. A whispering campaign grew, and the whisper had a name at the end of it.

The name did not ruin her. It made her. Nixon went on to concert halls and television and, much later, to a credited place in a Disney film — her singing contribution to Mulan (1998) carries her name. A voice that had been kept out of the credits for decades ended up being, in the end, on them.

And Bernstein belongs in the story as more than a composer who admired the result. He put a share of his income behind the singer. That did not put her name into the film's credits. It did put some of the soundtrack's earnings into her hands.

The name and the money are related. They are not interchangeable. You can win one and still have to ask for the other.

It would be easy to call this a story about one studio's bad faith, and so it is. But the larger fact is that a whole industry kept a secret on behalf of its customers — on behalf of people who would have preferred to be told. They were not protected. They were managed.

The silence was never hers.

That is how Michael Greene of the Recording Academy explained, on 19 November 1990, why Milli Vanilli's Grammy for Best New Artist was being taken back. Ten words. Not a word about whether the songs were good, or whether the audience had been robbed of a performance. A falsification of label credit.

It was a line on the sleeve.

Which is where this Feature began. Piece 02 was about a record that carried a character's name and not the singer's. Here is the other end of the same wire: a record that carried two men's names and not the people who had sung, and a correction delivered by a trophy committee.

Rob Pilatus and Fabrice Morvan — Rob and Fab — were two young performers recruited by the German producer Frank Farian. The album was Girl You Know It's True. It sold in enormous numbers: five top-five singles in the United States, three of them number ones, and the pair's faces on the cover and in the videos and on the stage. Their voices were not on the record.

The voices belonged to other singers. Contemporary Associated Press reporting names John Davis, Charles Shaw and Brad Howell among them. The men whose faces sold the album and the men whose voices were on it were different men.

The slip itself belongs to the live show, where tracks skip. At a 1989 concert in Bristol, Connecticut, the recording of "Girl You Know It's True" caught on a repeated phrase. Contemporary reporting in November 1990 identifies Bristol as the scene. A voice went round in a circle. The men on stage could not sing their way out of it.

But the tape did not end the arrangement that night. In February 1990 they won the Grammy. In November, Farian admitted that other people had sung on the record. On 19 November the Academy revoked the award.

A year separated the visible failure of a backing track from the admission about the album. The suspicion came first. The Grammy came next. The confirmation came last.

Now look at the order of things, because it is the same order as piece 02.

Nobody objected to the sound. The sound had been the point. Three number ones is not the arithmetic of a product that nobody wanted. People bought the album because they liked the songs. They sang them in cars. The songs did not become worse in November 1990. They became, retroactively, a different kind of object — one where the credit and the voice had been separated by the people selling them.

So the audience's anger was not about the voice. It was about the information.

And here there was a question about where the award should go. Pilatus and Morvan said they wanted it given to the real singers. The Associated Press report records that request, and says the Academy had not yet decided whether to give it to another artist. Revoking a false credit and recognising the people who did the work were two separate decisions. The first could be made by taking something away. The second required deciding who should receive it.

Which is the second thing the story shows: what a credit is for. A credit is not a courtesy. It is how the money and the awards find their way to the right house. When the credit is false, the trophy goes to the wrong house. Taking it back corrects the award. It does not, by itself, give the singers their due.

It cost the faces a great deal. Within the year, the two men were famous for the one thing they had been hired to hide.

There is a version of this story that is about vanity — two young men who loved the fame. It is a bad version. Whatever Rob and Fab wanted, they were in an arrangement with a producer who controlled the voices, the image, and the story. The one thing nobody in the arrangement controlled was the audience, which had been told one thing, and found out another.

Compare Marni Nixon in piece 03. Her account is of a studio enforcing silence. Here the record itself supplied a false answer. Neither arrangement gave the listener a reliable way to connect the voice with the person who had made it.

A secret is a bet that nobody will check. The Bristol tape made that bet harder to keep. It did not stop the Grammy arriving months later.

My reading is that one honest line on the sleeve — naming the people who actually sang — would have changed the terms on which the record was sold. Whether it would also have saved the Grammy is a different question.

The sound had not changed. What the audience knew about it had.

She was seven. A recording of her singing "Ode to the Motherland" played at the opening ceremony of the Beijing Olympics, on 8 August 2008, as the Chinese flag was carried into the stadium. The stadium heard her. The child on view was someone else.

What they saw was Lin Miaoke — a pigtailed girl in a bright red dress, standing in the middle of the stadium, mouthing the words.

It took four days for anyone to say so. On Tuesday, 12 August, the ceremony's general music designer, Chen Qigang, a contemporary composer and a French citizen, gave an interview to a state broadcaster, and said it plainly: the voice was Yang's, the face was Lin's. The reason he gave is the sentence that carried the story round the world. Yang, he said, was not chosen to appear because "we wanted to project the right image."

Compare that with piece 03. Nixon's studio had a contract and a threat. Beijing had an interview and a rationale.

How did it come about? Chen's own account is that a senior member of the Communist Party's Politburo attended a rehearsal and raised a concern. The concern, as Chen tells it, was that there was a problem, and that it needed fixing. The official was not named in the contemporary reports of Chen's account. He also said the decision had been made with the national interest in mind — "what was best for the nation," in the report's paraphrase.

Hold on to the ingredients. A rehearsal. A powerful person in the stalls. A problem. And a solution: one child for the ear, a different child for the eye.

The press reports of the day added a detail that was not Chen's. Yang Peiyi, the voice, was described as having a chubby face and uneven teeth. Those are descriptions in the press reports, rather than the words quoted from Chen here. It's the kind of detail that sticks to a story because it explains itself too well. But it is a reporter's observation, and it should be read as one.

Chen, for his part, called Yang's voice "perfect" and said that everyone had agreed. He also said Lin was excellent "in those aspects," meaning, as I read it, the aspects of appearance and presence. Photographs of Lin in her red dress had gone round the world, and the official China Daily had called her a rising star. Her face was on the story before the story was true.

The mechanism was different from Nixon's account of contractual silence, and different again from an album sold under false vocal credits. A national spectacle was staged. In Chen's account, an intervention from above the composer helped decide which child the nation would hear and which it would be shown. That is what the reports establish; they do not establish all the arrangements made with the children or their families.

And for all that, it is the same arithmetic. The voice and the face had been separated by someone who thought the audience would prefer it that way. The audience, once told, did not.

There is, though, a thing in this one that neither of the others have.

In Hollywood, in Bombay, in Munich, the ghost was kept from the audience in order to sell something to it. In Beijing the ghost was kept from the audience in order to sell it itself. The product was a nation. A patriotic song accompanying the arrival of its flag was heard in the recorded voice of one child and mouthed by another. (China's national anthem is March of the Volunteers, a different song.) Chen's account places the intervention at a rehearsal, with a member of the body that governs the country. The choice of a face was being made in the name of the nation.

That is a different kind of credit. It is credit as a political fact.

(Last month's Feature, Opening Nights, was about audiences who fought over a play. Here was an Olympic audience, handed nothing to fight about.)

Chen Qigang said afterwards that the child with the face had been chosen for her cuteness and the child with the voice had been chosen for her voice. Both of those statements are true. The difficulty is the verb at the front of them: chosen, by someone, about someone who was seven.

Seven years old. The account explains what the adults wanted to show the world. What Yang wanted is not answered by that explanation.

The distinction matters. Accounts of Lata Mangeshkar's demand describe a half-share of a five-per-cent royalty available to certain composers: two and a half per cent for the singers. Five per cent and two and a half are not necessarily rival versions of one figure. One can be the share from which the other is being asked.

In a 2014 interview reproduced by the Music Composers Association of India, Lata herself describes receiving a two-and-a-half-per-cent royalty out of a producer's share. That is evidence of an arrangement she obtained. It does not establish identical terms for every singer, every producer or every year of her campaign.

The dates are less tidy. Press Trust of India's account of Yasmin Khalid Rafi's memoir places the break at the beginning of 1960 and says it lasted two or three years. An Indian Express obituary gives 1963–67. These accounts disagree. The disagreement should travel with their names attached.

By this point in the Feature, the money has already entered the story. Bernstein gave Nixon a share of his royalty. Here the argument was whether a singer should have to depend on someone agreeing to give it.

Begin with what is not in dispute. Lata Mangeshkar and Mohammed Rafi were two of Hindi cinema's defining duet voices: the heroine's voice and the hero's, over and over. By the early 1960s both were named. Piece 02 followed the credit as it caught up with her. Nobody in this quarrel was a ghost. The credit was on the record.

The question was what came with the credit.

The arrangement Lata challenged was the session fee as the end of the singer's claim. Nixon, in piece 03, had met the same boundary when producers refused her request for royalties. A fee paid for work done. Nothing in that fee, by itself, guaranteed a share of what came afterwards. The record went on selling. The film went on playing. The song went on being sung at weddings. The singer had been paid for a day's work.

Lata's position, in the accounts that agree about it, was that this was not enough: that singers who made a record sell should share in what it earned. Rafi's, as Yasmin Khalid Rafi's memoir is quoted in the PTI report, was that a singer's claim ended when the producer paid the agreed fee. You could put each of those on a placard. Neither is crazy. They are two ways of answering the question of what a voice is: a service, rendered once, or a share in something that goes on.

So they stopped singing together.

The reconciliation has acquired several custodians too. The PTI account of the memoir credits Jaikishen. Shahid Rafi, interviewed in 2011, suggests Kalyanji–Anandji rather than S.D. Burman. They did sing together again. Who gets the credit for bringing them back is a less settled story.

But a reconciliation between two singers could not settle the terms on which an industry paid them.

Rafi died in 1980. Lata lived until 2022, long enough to see the argument enter the law and to take part in the organisation built to collect the money.

In 2012 Parliament passed the Copyright (Amendment) Act. The amended Section 38A gives performers exclusive rights over specified uses of their performances, and expressly provides for royalties for commercial use even where they have consented in writing to a performance being incorporated in a film. It is a legal framework, not a retrospective verdict on every term of the Lata–Rafi quarrel. It does not set the two-and-a-half-per-cent rate discussed here.

The singers then organised to collect. The Indian Singers' Rights Association was incorporated on 3 May 2013. The society, now called the Indian Singers' and Musicians' Rights Association, names Lata among the people who formed it, alongside Usha Mangeshkar, Suresh Wadkar, Alka Yagnik, Sonu Nigam, Sanjay Tandon and others.

She was not simply a name from the beginning of the story. She was there when the machinery was being built.

Fifty years, give or take. A quarrel in a recording studio in Bombay, and a registered society in 2013.

Now put the six pieces together. In Calcutta in 1935 the recording came before the camera, in an experiment that helped establish playback. In Bombay in 1949 a voice was filed under a character, and a singer's demands and listeners' questions challenged the omission. In Hollywood the voice was concealed for the sake of a product, while a share of its earnings reached Nixon through Bernstein. Milli Vanilli's record sold under a false vocal credit; a failing tape raised suspicions, and a producer's later admission preceded the loss of the Grammy. In Beijing a voice was concealed for a national image, and a composer's interview explained the substitution.

The name and the money keep crossing. Neither waits politely for the other to be settled.

But in the Bombay quarrel, the names were already known. Lata and Rafi were not asking to be identified. They disagreed about what followed from being the people who sang.

Not only who sang this. But who is owed.

The title of this piece is a fraction. The distance it covers is half a century: from asking for a share to building a society that could collect one.

A name can go on a label in a line. A right needs machinery.

These stories reach us by different routes: a singer remembering a studio, a collector describing two pressings, a reporter recording what an awards committee said, a statute, a society's account of its formation. They do not all answer the same questions. The Kamini label and the precise date it changed are different questions. So are the existence of the Lata–Rafi break and the year it began. Where the accounts disagree, we have kept the disagreement attached to the people who give it.

Our reading of what these episodes mean is an argument. The names, dates and payments underneath it need sources, and those sources can change the argument. A gap in our research is not proof that a record never existed. The links in these pieces are a way back to the accounts we have used, and an invitation to bring a better one.

That is part of keeping a record too. Leaving it open to correction.