In May, police officers in the Bronx shot 25-year-old Rameek Smith 19 times. He died, officials said, from gunshot wounds to the head.

The NYPD said Smith had fired first, hitting a police officer in the arm.

Nearly four months later, on the Friday before Labor Day, the NYPD posted footage on YouTube from two officers’ body-worn cameras that depict the minute leading up to the shooting. Soon after, Gothamist published a story with a link to the video, and described the images as failing to clearly show what cops said had happened: Smith pulling a weapon on them first, and squeezing off three rounds. Our story said the tape does plainly show the officers shooting at Smith, however.

Critics soon started posting on social media that Gothamist had gotten the story wrong. They downloaded the video and posted screenshots that they said showed Smith shooting first. So we did more reporting to find out what experts say the video shows, and doesn’t show.

A New York-based audio and visual forensic technician did an analysis of the video for Gothamist. After running the video through software to enhance the image, stabilize the recording, and slow it down, the technician, Al Zlogar, examined the video frame by frame. He found that Smith indeed fired two or three rounds before Officers Dennis Vargas and John Echevarria returned fire.

“Stabilizing it 100% and playing it back frame-by-frame, and knowing what you’re looking for, helped me finally see it,” Zlogar said. He said the process took him about a half hour.

And yet most people watching such videos don’t have the tools and background to do such an examination, which partly explains the public disagreement about what the images depict. According to those with professional experience reviewing body-worn camera footage, the NYPD’s recording of the Smith shooting exposes the flawed promise of body-worn cameras as a panacea of police transparency.

After the video was released, Peter Moskos, a professor at John Jay College of Criminal Justice and a former police officer, downloaded the video, zoomed in, and slowed it down. Moskos, who has a doctorate in sociology but is not a forensic video analyst, said he saw a muzzle flash, smoke, and the victim’s weapon.

One forensic expert, Ed Primeau, looked at Moskos’ screenshot and told Gothamist that the image was too over-enhanced to make a definitive determination. Regardless, the varied reactions to the analysis of the police footage proved a larger point that Moskos made in a follow-up interview: The way one views body-worn camera footage is reflective of how they view cops.

As for Moskos, he believes police accounts are “usually true,” and “I assume the guy did shoot at cops, because that’s what they said.”

Others who watched the video disagreed, saying they did not see Smith fire. In the end, what was clear is that videotaped evidence, often thought of as definitive proof of what happened during a disputed set of events, is not always straightforward.

The official investigation is ongoing

The state attorney general’s office is investigating the shooting, and its video could include more footage. Last year, police officers killed Eudes Pierre in Brooklyn, and body-worn camera video released by state Attorney General Letitia James’ office included bodycam footage of the aftermath of the shooting — including officers handcuffing the victim and calling for an ambulance — that was not in the video released by the NYPD.

The attorney general released that video within three months of the shooting, but no similar footage has been released by James' office nearly four months after Smith’s killing. Her spokesperson did not explain the delay.

The NYPD’s force investigation division will also review the evidence to determine whether police followed department protocols.

In the immediate aftermath of the Smith shooting, police said two officers from Mayor Eric Adams’ new anti-gun units, called Neighborhood Safety Teams, shot Smith in the head because he had fired on them first. They weren’t wearing the regular police uniforms, and they rolled up to him in an unmarked car at about 10:45 p.m. on May 10. Adams had said the incident was another case of criminal justice reform going too far, because Smith had an outstanding gun charge from 2020 and he was out of jail awaiting sentencing. And police said he was illegally carrying a stolen gun that night.

No further information on the case was released until the Friday before Labor Day, without a press release. Spokespeople for the NYPD didn’t answer a question about why it took four months for the video to come out (NYPD rules say it should be within 30 days for a fatality, with a few exceptions).

The timing of the video’s release raised the suspicions of Rebecca Kavanagh, a criminal defense attorney and legal analyst whose viral Twitter thread first cast doubt on the assertion that the video backed up the police version of events.

“Body cameras are always sold as increasing transparency and accountability,” Kavanagh said in a follow-up email. “But that is completely undermined when the city chooses to dump the footage late in the afternoon on the Friday before Labor Day and release a selected portion.”

The body-worn camera footage did not show what happened after Smith was shot. Police said they recovered a gun that Smith allegedly used, but it is not seen in the tape and it is not clear if the camera was turned off at this point, which would violate NYPD rules.

It's possible the footage was withheld from the YouTube post. The NYPD did not respond to questions about this footage.

Seth Stoughton, a law professor at the University of South Carolina who runs training sessions for law enforcement agencies, defense attorneys, prosecutors, and judges on body-worn camera analysis, said police departments have inconsistent practices on when they release footage, which “can undermine the public trust.” And while some law enforcement agencies include explanation and annotation when they release a video, others don’t.

In the Smith case, the NYPD did not isolate the image that they say shows Smith firing, nor did they use an arrow to point to where Smith’s weapon may have been, unlike other agencies do and unlike in at least one past case in New York.

“In the era that we currently live in, we tend to assume, without even thinking about it, that video is a more objective, comprehensive, and accurate source of evidence than any other piece of evidence,” Stoughton said. “So sometimes when we watch video, we assume we have all the context we need, right? ‘The video is in and of itself complete; I don’t need any other context to understand what’s going on in the video.’ And that’s not true.”

Same video, seen different ways

Part of the problem is the way our brains work, Stoughton said. The field of vision on body-worn camera footage is narrower than the human eye, but people tend to assume what they see on video is the entirety of what was experienced. An officer, for example, could be reacting to what he sees over his shoulder — even though the body-worn camera is facing straight ahead.

“And the scary part is we’re not even aware, consciously, that we’re filling gaps,” Stoughton said. “We are always making assumptions about things without even being aware that we’re making assumptions.”

For example, body-worn camera footage’s mic-against-the-uniform sound and shaky movement can make a scene look more chaotic and dangerous than it really was, and therefore serve to favor the police perspective, he said.

There are also irregularities in how body-worn cameras are activated. NYPD rules say officers must turn on their cameras “prior to engaging in” police actions, including “interactions with persons suspected of criminal activity.”

But the lack of audio in the Smith video — which could shed critical light on how the incident began and whether Smith realized it was officers who were chasing after him — appears to be due to the fact that the camera wasn’t turned on when it was supposed to be. That’s because when an officer presses the button to turn the camera on, it automatically buffers back to record at least 30 seconds of video preceding activation — yet that buffered recording does not include sound, according to NYPD guidelines.

In this case, the button doesn’t appear to have been pressed until after the foot-chase began and just before the officer opened fire.

A similar delay in activating body-worn cameras appears to have occurred during a traffic stop in March, in the same precinct in the Bronx. Officers riding in an unmarked vehicle shot into the vehicle of 18-year-old Luis Monsanto, but there’s no audio on the body-worn camera footage until after the non-fatal shooting. Adams said afterward that Monsanto tried to flee by driving at the officers.

Like Smith, Monsanto was shot in the head. Last week, Monsanto, who suffered a traumatic brain injury, sued the city and the officers involved.

All NYPD officers are equipped with body-worn cameras. The agency said it has the largest body-worn camera program in the country.

Given the fraught debate over policing, the public has demanded to see all images from these cameras — and that’s something law enforcement agencies are grappling with, according to Jeffrey Bellin, a former prosecutor and professor at William & Mary Law School. He said police departments used to withhold raw materials from investigations, “and that’s something the public no longer accepts.”

“They don’t just want to rely on the NYPD or prosecutors to tell them what happened,” he said. “And now that that’s worked to a degree, that the agencies are letting us see the footage for ourselves, we’re running into some of these problems of we’re not experts on how to parse a view. So people are going to make mistakes interpreting the video.”

Correction: A previous version of this story misstated the university where Seth Stoughton is a professor. He is at the University of South Carolina.