WASHINGTON -- After contradictory stories emerged about an F.B.I. agent's killing last month of a Chechen man in Orlando, Fla., who was being questioned over ties to the Boston Marathon bombing suspects, the bureau reassured the public that it would clear up the murky episode.
"The F.B.I. takes very seriously any shooting incidents involving our agents, and as such we have an effective, time-tested process for addressing them internally," a bureau spokesman said.
But if such internal investigations are time-tested, their outcomes are also predictable: from 1993 to early 2011, F.B.I. agents fatally shot about 70 "subjects" and wounded about 80 others -- and every one of those episodes was deemed justified, according to interviews and internal F.B.I. records obtained by The New York Times through a Freedom of Information Act lawsuit.
The last two years have followed the same pattern: an F.B.I. spokesman said that since 2011, there had been no findings of improper intentional shootings.
In most of the shootings, the F.B.I.'s internal investigation was the only official inquiry. In the Orlando case, for example, there have been conflicting accounts about basic facts like whether the Chechen man, Ibragim Todashev, attacked an agent with a knife, was unarmed or was brandishing a metal pole. But Orlando homicide detectives are not independently investigating what happened.
"We had nothing to do with it," said Sgt. Jim Young, an Orlando police spokesman. "It's a federal matter, and we're deferring everything to the F.B.I."
Occasionally, the F.B.I. does discipline an agent. Out of 289 deliberate shootings covered by the documents, many of which left no one wounded, five were deemed to be "bad shoots," in agents' parlance -- encounters that did not comply with the bureau's policy, which allows deadly force if agents fear that their lives or those of fellow agents are in danger. A typical punishment involved adding letters of censure to agents' files. But in none of the five cases did a bullet hit anyone.
Critics say the fact that for at least two decades no agent has been disciplined for any instance of deliberately shooting someone raises questions about the credibility of the bureau's internal investigations. Samuel Walker, a professor of criminal justice at the University of Nebraska Omaha who studies internal law enforcement investigations, called the bureau's conclusions about cases of improper shootings "suspiciously low."
Current and former F.B.I. officials defended the bureau's handling of shootings, arguing that the scant findings of improper behavior were attributable to several factors. Agents tend to be older, more experienced and better trained than city police officers. And they generally are involved only in planned operations and tend to go in with "overwhelming presence," minimizing the chaos that can lead to shooting the wrong people, said Tim Murphy, a former deputy director of the F.B.I. who conducted some investigations of shootings over his 23-year career.
The F.B.I.'s shootings range from episodes so obscure that they attract no news media attention to high-profile cases like the 2009 killing of an imam in a Detroit-area warehouse that is the subject of a lawsuit alleging a cover-up, and a 2002 shooting in Maryland in which the bureau paid $1.3 million to a victim and yet, the records show, deemed the shooting to have been justified.
With rare exceptions -- like suicides -- whenever an agent fires his weapon outside of training, a team of agents from the F.B.I.'s Inspection Division, sometimes with a liaison from the local police, compiles a report reconstructing what happened. This "shooting incident review team" interviews witnesses and studies medical, ballistics and autopsy reports, eventually producing a narrative. Such reports typically do not include whether an agent had been involved in any previous shootings, because they focus only on the episode in question, officials said.
That narrative, along with binders of supporting information, is then submitted to a "shooting incident review group" -- a panel of high-level F.B.I. officials in Washington. The panel produces its own narrative as part of a report saying whether the shooting complied with bureau policy -- and recommends what discipline to mete out if it did not -- along with any broader observations about "lessons learned" to change training or procedures.
F.B.I. officials stressed that their shooting reviews were carried out under the oversight of both the Justice Department's inspector general and the Civil Rights Division, and that local prosecutors have the authority to bring charges.
The 2,200 pages of records obtained by The Times include an internal F.B.I. study that compiled shooting episode statistics over a 17-year period, as well as a collection of individual narratives of intentional shootings from 1993 to early 2011. Gunfire was exchanged in 58 such episodes; 9 law enforcement officials died, and 38 were wounded.
The five "bad shoots" included cases in which an agent fired a warning shot after feeling threatened by a group of men, an agent fired at a weapon lying on the ground to disable it during an arrest, and two agents fired their weapons while chasing fugitives but hit no one. In another case, an agent fired at a safe during a demonstration, and ricocheting material caused minor cuts in a crowd of onlookers.
Four of the cases were in the mid-1990s, and the fifth was in 2003.
In many cases, the accuracy of the F.B.I. narrative is difficult to evaluate because no independent alternative report has been produced. As part of the reporting for this article, the F.B.I. voluntarily made available a list of shootings since 2007 that gave rise to lawsuits, but it was rare for any such case to have led to a full report by an independent authority.
Occasionally, however, there were alternative reviews. One, involving a March 2002 episode in which an agent shot an innocent Maryland man in the head after mistaking him for a bank robbery suspect, offers a case study in how the nuances of an F.B.I. official narrative can come under scrutiny.
In that episode, agents thought that the suspect would be riding in a car driven by his sister and wearing a white baseball cap. An innocent man, Joseph Schultz, then 20, happened to cross their path, wearing a white cap and being driven by his girlfriend. Moments after F.B.I. agents carrying rifles pulled their car over and surrounded it, Agent Christopher Braga shot Mr. Schultz in the jaw. He later underwent facial reconstruction surgery, and in 2007 the bureau paid $1.3 million to settle a lawsuit.
The internal review, however, deemed it a good shoot. In the F.B.I.'s narrative, Agent Braga says that he shouted "show me your hands," but that Mr. Schultz instead reached toward his waist, so Agent Braga fired "to eliminate the threat." While one member of the review group said that "after reading the materials provided, he could not visualize the presence of 'imminent danger' to law enforcement officers," the rest of the group voted to find the shooting justified, citing the "totality of the circumstances surrounding the incident," including that it involved a "high-risk stop."
But an Anne Arundel County police detective prepared an independent report about the episode, and a lawyer for Mr. Schultz, Arnold Weiner, conducted a further investigation for the lawsuit. Both raised several subtle but important differences.
For example, the F.B.I. narrative describes a lengthy chase of Mr. Schultz's car after agents turned on their siren at an intersection, bolstering an impression that it was reasonable for Agent Braga to fear that Mr. Schultz was a dangerous fugitive. The narrative spends a full page describing this moment in great detail, saying that the car "rapidly accelerated" and that one agent shouted for it to stop "over and over again." It cites another agent as estimating that the car stopped "approximately 100 yards" from the intersection.
By contrast, the police report describes this moment in a short, skeptical paragraph. Noting that agents said they had thought the car was fleeing, it points out that the car "was, however, in a merge lane and would need to accelerate to enter traffic." Moreover, a crash reconstruction specialist hired for the lawsuit estimated that the car had reached a maximum speed of 12 miles per hour, and an F.B.I. sketch, obtained in the lawsuit, put broken glass from a car window 142 feet 8 inches from the intersection.
The F.B.I. narrative does not cite Mr. Schultz's statement and omits that a crucial fact was disputed: how Mr. Schultz had moved in the car. In a 2003 sworn statement, Agent Braga said that Mr. Schultz "turned to his left, towards the middle of the car, and reached down." But Mr. Schultz insisted that he had instead reached toward the car door on his right because he had been listening to another agent who was simultaneously shouting "open the door."
A former F.B.I. agent, hired to write a report analyzing the episode for the plaintiffs, concluded that "no reasonable F.B.I. agent in Braga's position would reasonably have believed that deadly force was justified." He also noted pointedly that Agent Braga had been involved in a previous shooting episode in 2000 that he portrayed as questionable, although it had been found to be justified by the F.B.I.'s internal review process.
Asked to comment on the case, a lawyer for Agent Braga, Andrew White, noted last week that a grand jury had declined to indict his client in the shooting.
In some cases, alternative official accounts for several other shootings dovetailed with internal F.B.I. narratives.
One involved the October 2009 death of Luqman Ameen Abdullah, a prayer leader at a Detroit-area mosque who was suspected of conspiring to sell stolen goods and was shot during a raid on a warehouse. The F.B.I. report says that Mr. Abdullah got down on the ground but kept his hands hidden, so a dog was unleashed to pull his arms into view. He then pulled out a gun and shot the dog, the report says, and he was in turn shot by four agents.
The Michigan chapter of the Council on American-Islamic Relations filed a lawsuit against the F.B.I. The group was concerned in part because the handgun had no recoverable fingerprints and because of facial injuries to Mr. Abdullah. It also contends that the dog may have been shot instead by the F.B.I. agents and the gun thrown down in a cover-up.
A report by the Michigan attorney general's office, however, detailed an array of evidence that it says "corroborates the statements of the agents as to the sequence of events," including that bullet fragments in the dog's corpse were consistent with the handgun, not the rifles used by the F.B.I. agents. Such an independent account of an F.B.I. shooting is rare. After the recent killing of Mr. Todashev in Orlando, both the Florida chapter of the same group and his father have called for investigators outside the F.B.I. to scrutinize the episode.
James J. Wedick, who spent 34 years at the bureau, said the F.B.I. should change its procedures for its own good.
"At the least, it is a perception issue, and over the years the bureau has had a deaf ear to it," he said. "But if you have a shooting that has a few more complicated factors and an ethnic issue, the bureau's image goes down the toilet if it doesn't investigate itself properly."
This article originally appeared in The New York Times.