For years after the assassination of Robert F. Kennedy in June 1968, analysis of the case focused on the ballistics evidence and varying witness accounts of what happened in the pantry of the Ambassador Hotel in Los Angeles. But in 2004, delving through the archives of the police case file, CNN senior writer Brad Johnson found a recording that had been ignored for decades.

It was a tape inadvertently made by Polish journalist Stanislaw Pruszynski, who had recorded Kennedy’s victory speech with a cassette recorder. Kennedy then left the stage and walked through the pantry while Pruszynski detached his microphone from the podium, video of the event shows. As Pruszynski then moved toward the pantry, his tape was rolling. When Johnson listened 36 years later, he thought he heard shots, then a scream.

No other recording of the shooting had ever been discovered. No cameras were rolling as Kennedy shook hands with bus boys in the pantry, followed by gunfire described by many witnesses as two shots, a pause, and then a rapid-fire sequence of shots. Johnson turned the tape over to audio engineer Philip Van Praag. Using modern technology to limit ambient noise and slow down the tape, Van Praag counted 13 shot “impulses,” or wave forms resembling gunshots, and possibly more drowned out by screams.

It was the first new discovery in the RFK case in years. If true, it was monumental – Sirhan Sirhan’s .22-caliber Iver Johnson Cadet revolver held eight shots. Police said those bullets were accounted for, between the three shots that struck Kennedy and the five that struck five people standing behind him. But more than eight shots meant a second gun was involved, as many people critical of the investigation had argued as early as 1969.

Though some experts disagree with Van Praag, his analysis has become a key piece of evidence for those demanding a new investigation into the assassination. Kennedy’s son Robert F. Kennedy Jr. recently joined that call, in part because “There were too many bullets,” he told The Washington Post. “You can’t fire 13 shots out of an eight-shot gun.” Kennedy’s daughter, Kathleen Kennedy Townsend, also endorsed the request.

After listening to the recording, some acoustic experts say Van Praag misidentified some sound impulses as gun shots and that only seven or eight shots could be heard on the low-fidelity tape. Forensic acoustics engineer Philip Harrison said, “I can’t find any more than the seven shots that are there.” In 2013, comments by Harrison and others helped convince U.S. Magistrate Judge Andrew J. Wistrich that the tape was no smoking gun. “Van Praag’s opinion is far from ‘conclusive’ evidence of a second gunman,” Wistrich wrote, “because other experts analyzing the Pruszynski recording have reached contrary conclusions.”

But Van Praag hadn’t stopped with simply counting the gunshot sounds. Then Van Praag revealed two additional findings: That two pairs of shots occurred so closely together that they couldn’t have been fired from the same gun, and then the finding that two guns were involved firing in opposite directions — one facing Pruszynski, presumably Sirhan, and one facing away from Pruszynski

Those who had been pushing since the early 1970s for a new investigation into a second gunman, such as wounded victim Paul Schrade, saw Van Praag’s discovery as having “special meaning based on the actual bullet count” of Sirhan’s eight-shot gun.

The Los Angeles police, the district attorney and the California attorney general’s office all declined to be interviewed about the assassination and consider the case closed. All of Sirhan’s appeals at the state and federal level were rejected, including one in 2013 that considered evidence not presented at his trial in 1969, including ballistics evidence, evidence that Sirhan may have been hypnotized and the Pruszynski tape.

In June 1968, Pruszynski returned to Montreal after the shooting, where he was then living. CNN’s Johnson, a longtime RFK investigator and the co-author of a new book on the case, “The Assassination of Robert F. Kennedy,” said Pruszynski did not reach out to authorities because he presumed other news media had captured the shooting with higher quality equipment than his portable cassette recorder.

But the FBI tracked down Pruszynski and arranged for the Royal Canadian Mounted Police to make a dub of his recording in about early 1969, as Sirhan’s trial was beginning in Los Angeles. An open-reel dub of the tape was sent to the FBI, which decided that there was nothing pertinent to the investigation on the tape, Johnson said. Van Praag said the technology was not available in 1969 to detect the gunshots.

The FBI provided a dub to the Los Angeles police, but it appears the recording was not provided to Sirhan’s defense team at any time, Johnson said.

After Sirhan’s trial ended in April 1969, the police refused to release any of their files for another 19 years, even defeating a lawsuit by Schrade. But city officials ordered the files released in 1987, and they were made available to the public in 1988, though Johnson said the recording itself wasn’t released until 1990, as archivists were going through all of the related audio in the files.

Johnson said it was 14 years to the day of the recording’s release that he sat listening to the Pruszynski tape and heard “popping sounds.” He compared it to film and video from the ballroom where Kennedy had just made his victory speech in the California Democratic primary and found the audio recording matched the moment when the shooting occurred, 12:16 a.m. on June 5. Mel Ayton, author of a book about Sirhan called “The Forgotten Terrorist,” said some experts have found that the times do not match.

Johnson said he ran the recording past sound technicians at CNN and Paramount Pictures, and they agreed there were more shots than could be heard by the naked ear. Johnson then recruited two audio specialists for deeper analysis: Van Praag, an audio engineer with extensive experience in the capabilities of magnetic tape recording; and Spencer Whitehead, an audio science professor at Georgia Institute of Technology. Both said more than eight shots had been captured on the Pruszynski tape. In 2007, three other experts who examined the recording for Johnson located “at least ten” shots, Johnson said.

“You can hear what sounds like it could be gunshots,” Van Praag said in an interview. “But it’s impossible to discern how many. … the first shots are really inaudible to the naked ear.” Witnesses said Sirhan fired two shots, then a pause, then a barrage of shots.

Using an oscilloscope and a spectrum analyzer, Van Praag said it’s possible to filter out the “gun” impulse sounds from other ambient sounds. He said a gunshot produces a very distinct impulse of extremely short duration, and that the trailing sound waves give it characteristics that differentiate gunshots from firecrackers or balloons popping. He said the police ShotSpotter technology works the same way, alerting police to gunfire but not to other, similar sounds. Harrison noted that in Birmingham, England, ShotSpotter produced so many false alerts that police removed it.

Van Praag said he found what appeared to be 13 shots, but a scream at the end of the shooting could have obscured one or more shots. Harrison and another opposing audio expert, Michael O’Dell, said in a 2011 declaration in Sirhan’s case that they found seven likely gunshot sounds and the scream may have obscured the eighth shot.

Harrison, who testifies in court cases for JP French and Associates, said that analyzing sound waves is “very much an interpretive exercise.” On the Pruszynski tape, “I’m not saying there’s nothing there. But his interpretation of gunshots, I’m saying I’m not convinced that there are” more than eight shots.

Looking more closely at the oscillograph, Van Praag found that shots three and four, and then shots seven and eight, are too close together to have been fired by the same gun. A documentary by the Discovery Times channel found that Sirhan’s .22-caliber pistol couldn’t have been fired that quickly. That further convinced Van Praag that a second gun was present in the hotel pantry. “Once you recognize there are more than eight,” Van Praag said, “then two guns were fired.”

Next, Van Praag used sophisticated software to determine subtle differences in the resonance characteristics of the gunshots, and fired different guns to analyze their sounds. He said he was able to show that shots 3, 5, 8, 10 and 12 were fired from a different gun, because their sound waves had an anomaly not found in the other eight shots. He said the five anomaly shots were fired in a different direction from the other eight shots. Van Praag said his tests showed those five shots were fired by an H&R 922 .22-caliber revolver — Sirhan was firing an Iver Johnson Cadet .22-caliber revolver — in the direction opposite of Sirhan’s gun.

“The conclusion is inescapable,” Van Praag wrote in a 2011 declaration in Sirhan’s case, “that there was a second gun fired by a second shooter during the shooting that resulted in the death of Senator Robert F. Kennedy, and that the five shots from the second gun were fired in a direction opposite the direction in which Sirhan fired.”

“I think it’s a flawed analysis,” Harrison said. He said Van Praag’s test shots were fired outdoors, but the Kennedy shots were fired in a cramped pantry. “Outdoor acoustics are very different from indoor,” Harrison said. Harrison produced his own oscillograph of the Pruszynski tape that showed seven shots and a possible eighth, also drowned out by screams, he said. Van Praag said doing a test outdoors proved the gun created the anomaly, not the indoor environment.

O’Dell filed a declaration in Sirhan’s case saying that he and Harrison both concluded that seven sounds “could safely be attributed to actual gunshots, with the eighth potentially masked by a loud scream” but that other “implosive or plosive sounds were different, and might have environmental causes other than gunshot.”

The California attorney general’s office noted that Sirhan admitted firing the shots that killed Kennedy, though he later said he did so at his lawyers’ instructions and that he didn’t remember firing the shots. The attorney general noted that “under California’s vicarious liability law, [Sirhan] would still be guilty of murder and assault even if there were a second gunman who actually shot Senator Kennedy.”

Johnson continued to push the case, and Robert Kennedy Jr. provided an assist in 2012 when he sent a letter to then-Attorney General Eric H. Holder supporting Schrade’s call for a new investigation. Johnson said Holder ordered the FBI to reexamine the Pruszynski recording and Van Praag’s findings.

In 2013, the FBI issued a Report of Examination that contradicted all the experts on both sides and said the tape was “of insufficient quality to definitively classify the impulse events as gunshots, confirm the number of gunshots or determine the identification of specific weapon(s).”

Van Praag said the FBI had used antiquated techniques and didn’t use technology with sufficient resolution to discern the resonance differences between individual gunshots. The FBI also never contacted Van Praag to discuss his findings.

Johnson said the “bureau’s examination was incomplete and its report was inconclusive,” and “doesn’t rise to the level of the FBI in doing serious work. They phoned it in without picking up the phone and calling Van Praag.” Both he and Van Praag said a complete, high-tech analysis of the Pruszynski tape is still needed to eliminate any questions about whether more than eight shots were fired in the Ambassador pantry.

Kennedy lived for about 25 hours after the shooting, and died 50 years ago today, at 1:44 a.m. on June 6, 1968. He was 42 years old.

More on the assassination of Robert F. Kennedy: