AJR  Features
From AJR,   April 2002

The Civilian Casualty Conundrum   

Have American news organizations soft-pedaled the “collateral damage” of the fighting in Afghanistan? Or have foreign news outlets and academic studies grossly inflated the toll?

By Lucinda Fleeson
Lucinda Fleeson is director of the Hubert H. Humphrey Fellowship Program at the University of Maryland. She has trained journalists in Eastern and Central Europe, Africa, Latin America and, most recently, Sri Lanka, where she was a Fulbright Scholar. Her training manual for teaching investigative reporting in developing democracies has been published in 18 languages by the International Center for Journalists.     


Throughout the evening of January 25, a dozen American reporters and photographers at the seedy Noor Jahan Hotel in Kandahar, Afghanistan, debated whether they should travel into Uruzgan province, a Taliban stronghold. The Pentagon had just reported that the United States had completed a successful mission there, killing 15 al Qaeda leaders and capturing more. But the reporters based in Kandahar began hearing from sources that the strike was a blunder. American journalists had reported sporadically on mistakes that killed innocent civilians. This might be the most flagrant example.

Into the night, the reporters worked their satellite phones, calling editors and sources. The trip's payoff could be big. But the risks loomed large.

Two days earlier, Daniel Pearl had been reported missing in Karachi, Pakistan. Eight journalists had already been killed in the Afghanistan countryside. "It seemed like an incredibly risky thing to do," recalls Kevin Whitelaw of U.S. News & World Report, who had just arrived in Kandahar. "There was no way to assess the risk. The Afghan government didn't know, the U.S. troops can't tell you. There were Taliban forces around, there might be al Qaeda hiding. You've got to figure that an American reporter would be quite a prize for them."

Should they go?

Obtaining accurate accounts of civilian deaths is one of the most difficult challenges of war reporting. Journalists must weigh conflicting information, exaggerations and lies as they constantly debate: How many sources do we need? How reliable are eyewitnesses, who might be in shock or have political agendas? What good are second-hand accounts?

Those journalistic obstacles have been magnified exponentially by Afghanistan's impenetrable terrain and near Stone Age conditions that demand a form of Extreme Reporting. War correspondents travel in armed convoys, live in primitive conditions and are exposed to incalculable risks from bandits, warring tribes, land mines and stray bombs.

Like blind men each feeling a different part of an elephant and unable to describe the whole beast, reporters filed searing reports of individual episodes of civilian deaths, but a total count remained elusive. "The Taliban was probably overestimating the numbers and the Pentagon probably understating the numbers," says Andy Mosher, the Washington Post's deputy foreign editor. "Somewhere in the middle was the truth."

Military and human rights analysts say that final, reliable counts of civilian casualties in Afghanistan will be much more difficult to ascertain than in Iraq following the Persian Gulf War or in the Balkans following the NATO bombing. "It's going to be far harder for Human Rights Watch and organizations like that to go in and do a reconstruction," says Rear Admiral Stephen Pietropaoli, the Navy's chief of information. "Taliban and non-Taliban people are indistinguishable in dress and appearance. On its best day, the Afghan government has never been centralized." Kosovo is the size of greater Los Angeles; Afghanistan is almost the size of Texas.

As journalists struggled to sort out the truth, the U.S. media were accused of being slow to report the American killing of innocents, while Web sites, newspapers and broadcasters around the world flashed headlines that American bombs claimed the lives of hundreds, even thousands, of civilians.

Critics charge that when American media did report civilian deaths, the stories were not played very prominently. Retired Boston University Professor Howard Zinn wrote in the February 11 issue of The Nation: "These reports have been mostly out of sight of the general public (indeed, virtually never reported on national television, where most Americans get their news), and so dispersed as to reinforce the idea that the bombing of civilians has been an infrequent event, a freak accident, an unfortunate mistake."

Even mainstream media observers noticed similar trends. Said Geneva Overholser, former Washington Post ombudsman and Des Moines Register editor, after perusing European Web sites, "It's a question of weight given to what sources, how the headlines were written, where the stories played." She wondered whether Americans suffered from a "home-team" syndrome, similar to what infected the British media during the Falklands War in the 1980s.

When she raised that question with returning war correspondents on a February 13 Brookings Institution panel, they responded vociferously that they were willing, if not eager, to report military screw-ups and civilian deaths but had higher reporting standards than the foreign media. "We were ready to leap," said Tom Squitieri of USA Today.

In the absence of hard data, two American academics released studies based largely on Internet research of news accounts, one estimating 1,000 to 1,300 civilian deaths as of January 1, the other 3,767 as of December 6. But the reports relied exclusively on the work of others and didn't distinguish between painstaking reconstruction of bombing gone awry and uncritical acceptance of second-hand reports. They ultimately resembled a global version of the game of telephone: a refugee says that eight villagers died in an attack; Taliban officials transmute that report into 30 casualties; that claim is posted on a Pakistani news agency's Web site; six other Web-based news services pick up the figure; a British newspaper includes the claim in its news of the day out of Afghanistan; an American professor counts up all of these reports over three months and concludes that 3,767 civilians had been killed.

By contrast, an on-the-scene reconstruction by the Associated Press released in February put the total at between 500 and 600.

The deliberations of the journalists holed up in the Noor Jahan Hotel in late January illustrate the challenges of reporting on the fighting in Afghanistan. Already by that evening, several reporters had wasted nearly 24 hours pursuing the report of a mission gone wrong, following erroneous leads to two locations. Now some of them thought they had zeroed in on the correct site, but was it worth the trip?

Across town from the Noor Jahan that night in a rented house, Steve Inskeep of National Public Radio and John Fullerton of Reuters debated the risks. "When the military comes out with something and says this is how it happened, part of your job as a war correspondent is to go see if it's true, see if they're lying," says Inskeep. But NPR editors told him not to go: Uruzgan was too far away; the trip was too expensive and too dangerous. Reuters editors also nixed the journey. One of the agency's photographers had already been killed in Afghanistan.

Craig S. Smith, normally the New York Times' Shanghai correspondent, had only been in Afghanistan 10 days. He was uncertain about venturing into unknown territory. When he heard that Jonathan S. Landay of Knight Ridder and Michael Ware of Time magazine were going, Smith decided, "I'm going, too."

The next morning, only three reporters, accompanied by photographers, set out for Uruzgan. While reporters can reach some locations in Afghanistan armed only with a notebook, this was going to require more. Landay and Ware had assembled a convoy of three four-wheel drive vehicles and a small army--six guards armed with AK-47 machine guns, a shoulder-carried rocket-propelled grenade and a light machine gun. Smith set off separately, with two vehicles and three guards toting AK-47s. In Tarin Kot, the local governor arranged for Smith to take a second truckload, with eight more heavily armed men. "I had RPGs [rocket-propelled grenades] and the whole kit and kaboodle," says Smith. "I was glad to have them."

The 150-mile journey took 12 hours along dirt tracks. They drove up river beds, sometimes dry and sometimes flooded. Roads cut into the mountainsides were often not much more than rutted trails along sheer drops. They passed numerous trucks destroyed by bombs, with new graves nearby. One truck had crashed down into the riverbed, its interior bright with fresh blood. Smith drove through the night. The other convoy stopped at a roadside hostel.

The reward for chancing the precarious trek was large. Landay and Smith revealed that U.S. Special Forces had stormed the town of Uruzgan in a midnight raid and shot apparently innocent men who had been conducting a disarmament drive for the provincial governor, some with their hands bound with plastic handcuffs. Another 27 soldiers, government workers and prisoners were captured.

The reporters talked with dozens of town residents and council members, who described how U.S. Special Forces arrived on foot in the dead of night and assaulted a school, government compound and jail, where prisoners held on local criminal charges slept. After troops left, an AC-130 flying gunship blasted the buildings.

The stories, while groundbreaking, did not receive spectacular play. Only a few Knight Ridder papers and subscribers picked up Landay's report that 21 anti-Taliban soldiers had been killed, and most ran it inside. On January 28, the New York Times played Smith's story reporting the deaths on page A6, despite lobbying by the paper's foreign desk to get it out front.

Back in Kandahar, NPR's Inskeep regretted that he hadn't gone. He interviewed Afghan officials who confirmed that not only had the U.S. attack killed the wrong people, but American officials had assured the Afghan government that such a mistake would never happen again. Inskeep's editors gave him the OK to go to Uruzgan.

Inskeep also found convincing evidence. When he stopped along the way in the provincial capital of Tarin Kot, the local governor showed him a bag with two sets of plastic handcuffs allegedly removed from dead bodies. Traveling through a snowstorm, the radio reporter reached Uruzgan, where local residents related how they had been gathered in a room and presented with $1,000 each in crisp U.S. $100 bills. The governor told residents the money was "from someone who wants to help you." Later, the governor acknowledged in a taped interview that Americans had provided the money and instructed the governor to apologize to residents. "We had the story a million different ways," says Inskeep. "We talked to so many people, and they all said the same thing."

Inskeep wanted a photo of the money. He went back to Tarin Kot. The reporter's translator insisted that for a foreigner to arrive unannounced after dark at a resident's house would be a breach of etiquette, so the translator went to the house and snapped what was soon dubbed "the money shot." It depicted a man holding up 10 $100 bills.

Inskeep spent that night in the local police station--he had the safest bed that the villagers could offer. He wrote his story, uploaded the digital photos and filed by sat phone for the February 4 "Morning Edition." NPR posted the photos on its Web site, and other news organizations picked them up, fueling the growing attention paid to the story.

U.S. military officials continued to deny that the shootings were an error, saying initially that maybe they hadn't killed al Qaeda forces but rather Taliban. Then the Pentagon acknowledged that some were not Taliban but criminals. Defense Secretary Donald H. Rumsfeld contended that the local forces fired first, which is possible, says Smith, seeing as U.S. Special Forces had arrived in a surprise night raid.

There had been other well-documented reports of civilian deaths, including a December 21 attack on a convoy of tribal elders traveling to the inauguration of the new interim Afghan government. But the Uruzgan incident seemed somehow different. In a bombing campaign civilian casualties might be accepted as tragic but unavoidable. This was an instance of foot soldiers storming in during the night and snatching innocent people, shooting others while they lay handcuffed. It also constituted an about-face by top military officials.

"One thing I definitely learned there," says Inskeep, "was that I could not put credence in most things if I could not see them. In order to really, really have the story, you just had to be there, talk to eyewitnesses, go to the scene, look over the evidence. Otherwise, in many instances, you just don't have anything."

Loren Jenkins, NPR's senior foreign editor and a former Newsweek and Washington Post foreign correspondent, says that Inskeep had followed an old adage in war reporting: "Only go with what you know yourself, for sure. I learned that lesson in Vietnam."

Another Vietnam lesson learned by both press and military was an understanding of the importance of civilian casualties. U.S. military leaders learned the hard way that they could not outpace the public's willingness to accept the human cost of conflict. Students of military theory frequently cite the views of 19th century Prussian military theorist Carl von Clausewitz, who coined the term "center of gravity" to define the core strength of a military force, its hub of power that propels it into war. American military, for instance, assessed Iraq's center of gravity as Saddam Hussein and his elite Republican Guard. The center of gravity in America, however, is usually believed to be the will of the people – often a function of citizens' acceptance of deaths of its own soldiers as well as the so-called collateral damage of civilian casualties.

As a result, billions have been spent to refine precision weaponry to improve their killing capability and reduce unintended casualties. "In assessing targets, if there is a possibility of significant civilian deaths, the military gain might be offset by the blowback, so they won't do it," says defense consultant Richard King, a retired Air Force colonel who worked on several of the post-Persian Gulf War after-action reports.

Such efforts have led Rumsfeld to say repeatedly that the Afghan campaign is the least bloody war in American history. Bombing in urban areas has been so precise, say reporters, that often one building with Taliban connections has been demolished while adjacent structures are left standing.

The Washington Post's William Branigin says he doesn't believe that civilian casualties in Afghanistan total in the thousands, at least in the early phase of the war. "It happened, bombs went astray. But it was pretty rare," he says. "At the front line, where I spent a lot of time, whenever planes passed overhead, Northern Alliance soldiers showed no concern at all that they were going to be bombed."

But after months of steady denials of bombing mistakes and clamped-down restrictions on press access, U.S. claims of low casualties have been met with skepticism. "I think there have not been very many civilian casualties," says NPR's Inskeep. "But when the Pentagon tells you there have not been civilian casualties, they have not given you enough access so that reporters can repeat that statement with confidence."

The raid on Uruzgan was an unusual instance in which reporters could reasonably ascertain what had happened.

The Taliban had ejected all journalists from Afghanistan during the early weeks of the bombing campaign, at a time when the Taliban claimed that 1,500 innocent Afghans had died. Many of those reports were largely discredited when reporters were later escorted to sites that contained little evidence.

After the Taliban closed the borders, AP Pakistan Bureau Chief Kathy Gannon and photographer Dimitri Messinis were the only Western journalists in Kabul for more than two weeks in October. They were busy enough trying to cover the bombing in the capital. "We wanted to get out to villages," says AP's deputy international editor, Nicolas Tatro, "but it was just too dangerous, too remote, too hard to get to."

When the American military landed in Afghanistan, it, too, sharply restricted press access. Into November and December, most reporters remained in the relatively safe urban harbors of Kabul and Kandahar. Journalists occasionally ventured into the dangerous countryside, but a comprehensive count of civilian deaths remained elusive. The Pentagon refused to provide a tally, maintaining that verification was impossible.

Determining civilian casualties was also complicated by cultural practices. Afghans bury their dead quickly, sometimes within hours after death. They don't want strangers to see women's bodies. The Taliban used civilians as shields, bivouacking next to humanitarian facilities.

Reporters tell horror stories of second-hand, even eyewitness, accounts that didn't check out.

One such case happened in early January, when a U.N. spokesman stood up at the daily press briefing in Kabul and announced that he had a reliable source with credible information about a December 29 bombing attack in the village of Niazi Qala in eastern Afghanistan. Scores of innocent villagers had been killed, he said.

Two days later, U.S. News & World Report's Whitelaw was in a market talking with lumber traders about road security. He recounts: "One guy came up and said, 'I've got a story.' The trader provided one or two details that echoed the U.N.'s report, but said that the Taliban commanders and ammunition had been in the village."

U.S. News printed both versions of the attack on Niazi Qala to illustrate the complexity of differentiating between civilian and Taliban deaths.

"Information was one of the hardest things about this assignment," says Whitelaw. "I'd talk to two different people within the same ministry, and they would tell me two different things. You have to do these stories right. You can't report what two villagers say. You have to do what the reporters at the Post and Times did, and really get to places like Uruzgan, talk to 50 people, talk to as many people as you can. Even then, accounts are riddled with maybes and suppositions."

Squitieri, who reported from Afghanistan for USA Today, says: "The fact that an allegation is raised by one person in a village and then channeled through the U.N. or a humanitarian organization does not mean we rush to write a provocative story. U.S. reporters often take more time to develop those stories than a lot of foreign media. Standards used by U.S. journalists are generally more rigorous than others."

The University of New Hampshire is located in Durham, a small rural college town of pizzerias and not much else. The modern box that houses the Whittemore School of Business & Economics is one of the few buildings that differ from the university's New England architecture of colonial brick and white columns. Associate professor Marc W. Herold's fourth-floor corner office would have a nice view, if its windows were not blocked by haphazard stacks of books and reports piled high on every surface, veering toward collapse. Herold sits in the middle of this chaos, in front of an electric typewriter on his desk. A friendly, graying man, he wears faded jeans and a maroon cotton turtleneck.

Herold joined the university in 1975 to teach economic development and has remained as a tenured associate professor. Beginning with a doctoral study that compiled a database of foreign investments of American companies, he has developed a specialty in exhuming material from archives and news reports, from which he extrapolates new conclusions.

After the U.S began its air campaign in Afghanistan, two California academics asked Herold to contribute an article on civilian deaths to a collection of essays on September 11 and the ensuing war. The project appealed to him, as he opposes the American bombing campaign, arguing that the capture of Osama bin Laden and al Qaeda should be treated as a police matter. "It is simply unacceptable," he wrote, "for civilians to be slaughtered.... There is no difference between the attacks on the WTC whose primary goal was the destruction of a symbol, and the U.S.-U.K. revenge coalition bombing of military targets located in populated urban areas. Both are criminal. Slaughter is slaughter." In his report, which can only be described as a left-wing polemic, he also accuses what he calls "the corporate media" of devoting "only sparse moments to the topic of civilian casualties, obeying the Bush-Pentagon directives."

His dossier consists of a spreadsheet, which lists daily reports of casualties, locations and the names of Web sites he used as sources.

Sitting at a computer screen for 10 to 12 hours a day, seven days a week, for 12 weeks, he worked directly from Web sites, scribbling handwritten notes on three-by-five cards.

While he says he also consulted newspapers and conducted telephone interviews, the study was largely based on Internet research. Sometimes he worked at a computer in his home, other times at the university library. Two graduate students typed his notes into a computer spreadsheet. "Unfortunately they are foreign speaking, so there were a lot of mistakes in the English," he says. Sometimes, Herold says, multiple spellings of town names caused confusion as to whether reports were focusing on the same or different events.

Herold says that he went over the students' work and that his January 4 operation for a detached retina was not a handicap. "In October, November and December I had perfect eyesight. I have an awful lot of experience in taking notes."

When I scrutinized his material for the bombing of three villages on October 27 through 29, I found many reports from the Pakistan News Service, which turned out to be short, rewritten stories from other news sources, with no bylines, simply attributed to "agency reports."

In order to be sure that I had searched the same news articles used by Herold in his report, I asked him if I could examine his source material. Initially the professor discouraged a visit, saying that he was recovering from his eye surgery. Then he said he was busy updating his research and answering media requests. Then he said his files were in disorder, piled up in his bedroom. Then he said he didn't have copies of the press reports he used to compile his report. "Sometimes I printed articles, sometimes not," he explained.

Herold's work has attracted critics. An Indiana University political science professor, Jeffrey C. Isaac, called his study biased and methodologically weak, and cited several instances where Herold's figures were unsubstantiated. For instance, Herold reported "the U.S. alternative media noted that during the first week of bombing, 400 Afghan civilians had been slaughtered." "The cited source," wrote Isaac, "is an opinion piece by an individual who writes--with no substantiation--that four hundred people were killed." Isaac also traced what Herold purported to be an eyewitness account of a truck bombing reported in Albalagh. Albalagh, Isaac wrote, is an "Islamic E-Journal," apparently produced in Karachi and filled with extremist rhetoric.

There are dozens of like examples.

Herold did provide his documentation for 12 days of bombing. He did not have his original files for those incidents; instead, he spent three days re-creating the research. He could not provide data for two incidents. For those he had relied on only one media source, the Pakistan Observer, whose online archives were no longer available for those dates.

A number of Herold's citations were from British or Australian newspapers. While some of these publications had correspondents in the region, an examination of the articles shows that often their reports repeated unconfirmed Taliban numbers. Other citations were to Pakistani publications such as the Frontier Post, which on October 27 reported: "U.S. warplanes dropped up to ten bombs on Afghanistan's capital overnight, killing seven civilians and terrifying residents cowering in their homes, witnesses and a ruling Taliban official said Friday."

"I've been criticized for using second- and third-hand sources," Herold acknowledges. "But isn't that what reporters do? They get on the phone, they talk to people, they search the Internet."

Herold dismisses the possibility that he is counting and recounting the same incidents from Web sites that trumpet figures without any grounding in fact. He cited some instances that later were confirmed by more credible reports. Even so, he has revised his figures downward, to between 3,000 and 3,400 deaths. "I know I'm right as to the magnitude of deaths," he says.

Carl Conetta, director of the Project on Defense Alternatives based in Cambridge, Massachusetts, published another Internet-based study in January that estimated there was "a high likelihood that 1,000 to 1,300 civilians were killed."

Conetta limited his Internet search to American and other Western news accounts. He counted the total number of civilian deaths reported from up close, by on-the-scene eyewitnesses, and then compared that figure with the total deaths reported by other sources, primarily refugees. Then he applied his own formula. "We sort of averaged a factor of error," he explains. "When these reports were made from distance, they tended to overestimate by a factor of four or five."

Has his method ever been used to estimate war casualties before? "No. It's typical induction theory."

Despite the flimsy evidence used by Herold and Conetta, their work helped prompt news organizations to make their own assessments. "A lot of fairly large numbers were being casually thrown around in various publications," says Tatro, the AP's deputy international editor. "We wanted to see if this was so, and put together what we did know."

The wire service had been one of the news outlets that disseminated civilian casualty figures promulgated by the Taliban and reported in the foreign press, and it felt some responsibility for trying to make its own determination. The AP explained in a February 11 story: "International news organizations, including AP, reported fast-mounting Taliban casualty claims, always with the caveat that Afghanistan was almost completely closed to the outside world and the figures could not be verified. Even so, various death tolls, confirmed or merely claimed, took on a life of their own once they were electronically indexed and archived. In some cases they served as the basis for academic research."

Laura King, one of the AP's special correspondents in Afghanistan, spent much of January poring over hospital records, visiting bombing sites and interviewing eyewitnesses and officials. She coordinated reports from a team of AP reporters in other parts of the country. Although the wire service was careful to warn that the final count was by no means definitive, the February 11 story concluded that the civilian death toll probably ranged from 500 to 600. In February the Boston Globe published a report using similar methods, estimating 1,000 civilian deaths.

King quoted Afghan journalists and officials who admitted their dispatches had been doctored. Mohammed Ismail, a reporter later promoted to acting director of the official Bakhtar news agency, said that in one typical instance he went to the site of an airstrike in Kabul's Khair Khana neighborhood on October 20 and saw eight bodies. "But it was changed in our dispatch to 20," he told the AP. When he heard the report on Taliban-run radio, the figure had soared to 30. Other Bakhtar journalists told King they had been ordered to report military deaths as civilian ones. She quoted reporter Younis Mihireen recalling a direct hit on a Taliban and al Qaeda housing complex that killed 60 fighters. "I saw it with my own eyes," said Mihireen. "There were no civilians anywhere nearby and I reported this. But the dispatch said all the dead people were civilians."

Tatro said that before King's report, "Many people had assumed this was the case. She proved it."

Without question, reports of civilian casualties have received more attention abroad than in the United States. The international press played such stories more prominently. They gave more credence to Taliban claims and were more skeptical of U.S. versions of battles and deaths, says Alice Chasan, editor of the New York-based World Press Review magazine, which publishes foreign media accounts.

"The Pakistani press would report that X number of people were killed in a strike, then the State Department would deny that any civilians were hit," Chasan says. "Then, pretty much routinely, the State Department would later say there had been in fact some casualties, but the number would never jibe with the Pakistani press." But that doesn't mean the latter was right. "The Pakistani press is very often inaccurate – inaccurate with precision in numbers and inaccurate with precision in facts," Chasan says.

Interviews with Pakistani journalists and those who have worked there confirm this view. Arnold Zeitlin, a former AP reporter in Pakistan in the early 1980s, returned there in 1998 to train local journalists. "I can say immediately that casualty figures reported in Pakistan and Indian news media are absolutely unreliable," he says.

Syndicated Pakistani columnist and former newspaper editor Mushahid Hussain also was skeptical about press reports coming out of his country, particularly dispatches that carry no bylines but are attributed to "agency reports." "Dubious credibility!" he warns. "Editorial controls are not in place."

Chasan says that trying to sift through stories to determine what is original reporting and what is reconstituted wire copy is a worldwide problem, not strictly a Pakistani one. "These are news organizations that operate on shoe-string budgets. Most American journalists don't realize the conditions under which most foreign reporters work," she says. Typically, reporters in developing countries are required to write several stories each day. Virtually none has budgets for American-style war reporting.

One Web site often cited by Professor Herold is the San Francisco-based Pakistan News Service.

Its Washington bureau chief, Syed Adeeb, is a pleasant, helpful man who explained that although PNS has foreign correspondents, much of its news is culled from other online news agencies. Most of the stories on civilian casualties were from another Web site, Information Times.

I thanked Adeeb, hung up the phone and checked out Information Times. I was surprised to find that its chief editor was none other than Syed Adeeb! I called him back for another chat.

Adeeb says that he receives reports from news services via e-mail and every day consults more than a dozen Web sites, mostly those of Pakistani news agencies. Do any of the agencies have reporters who actually go into Afghanistan to cover the war? I asked.

"Their editors don't allow it; it is very dangerous. They could get killed," he replied. Occasionally some reporters manage to get inside Afghanistan, he says, but "99 percent of the reporters are based in Pakistan. I haven't talked to any reporters who said they can actually tell me they were there."

Adeeb offered to send me a list of the Web sites he regularly visits. And, he added, "We also published several articles by an American professor who has studied the issue."

What's his name?

Marc Herold of New Hampshire.

Compiling the definitive number of civilian deaths may take months after the fighting stops, if it can be done at all. Two Afghan nongovernmental groups are undertaking a count. The New York-based Human Rights Watch sent a team of investigators in March. "By just physically examining the site we can tell an awful lot of what happened--what weapon was used, often what the intended target was," says Kenneth Roth, Human Rights Watch's executive director. "We try to find and speak with as many eyewitnesses as we can, both to determine who was killed [and] to find out whether any enemy troops were in the vicinity. We try to speak to people individually, alone, then cross-check their testimony."

Human Rights Watch senior military adviser William M. Arkin, who heads the team, is probably the world's leading expert on civilian casualties. "I've been to more targets that have been bombed than anyone in the U.S. Air Force or intelligence community," he says. "So don't fucking tell me we don't know what happens when bombs drop."

Arkin initially seems to fit the stereotype of the computer nerd, with a serious, bland appearance that masks a volatile intensity. But it's what's inside his Fujitsu laptop that is the real show. Flipping along on his high-speed T-1 computer connection at the Paul H. Nitze School of Advanced International Studies at Johns Hopkins' Washington, D.C., campus where he is a senior fellow, Arkin shows gruesome photographs of children and adults killed by NATO bombings in Serbia and Kosovo. "I have autopsy and death pictures of every single civilian who died in the bombings in Yugoslavia," he says.

During the three-month NATO air campaign in 1999, local media representatives and Yugoslav officials photographed bodies with digital cameras and e-mailed the evidence to Arkin. A Serb-speaking Human Rights Watch staff member investigated reports on the spot. When the war was over, Arkin and a team of Human Rights Watch workers traveled to Balkans battlefields and towns, in all visiting about two-thirds of the 90 sites where civilians died. They searched records and interviewed residents as well as Belgrade and provincial town officials. Their work resulted in a report, accepted as the most complete work, that said 500 civilians were killed in the NATO bombings. He performed a similar study after the Persian Gulf War.

A vocal critic of the military, Arkin hopes to show those who determine bombing targets how to avoid civilian casualties. "He is the best at this," says King, the retired Air Force colonel, who often relied on Arkin's data for post-gulf war reports. "He's a funny guy," continues King. "He's trusted by both sides of the ideological spectrum. When he was at Greenpeace [as director of military research], we didn't like his politics, but never doubted his credibility or analysis."

Arkin is a former Army intelligence analyst who writes military affairs columns for the Los Angeles Times and washingtonpost.com, appears on MSNBC and NBC and is an adjunct professor at the U.S. Air Force's School of Advanced Airpower Studies. He wrote a manual on how to use the Internet for research on military affairs.

His preparations for the Human Rights Watch expedition to Afghanistan took months. Step one was to assemble a comprehensive list of media accounts. Using the Herold material as well as his own collection of press reports, he identified about 350 reported incidents of civilian casualties and damage to civilian facilities.

Using software developed with a partner, he has built probably the most comprehensive publicly available maps of Afghanistan. He uses limited access military maps and satellite photographs, overlaid with his own reconstruction of the nation's roads, pipelines, runways, utility plants, houses of worship--all derived from public sources. Push a button, and the laptop screen is overlaid with purple splotches indicating population densities. He zooms in or out, focusing on a map showing just about every village in Afghanistan. Select another program, and the map shows all the sites of media accounts of casualties, marked with blue stars. Place the cursor on any star, and get a pull-down menu indexing a chronology of bomb reports for that site.

In step two, Arkin entered the 480 fixed targets where the U.S. thus far had dropped bombs during the air campaign, derived from a 10-page military document, arranged by date. "I just got it, all right?" he says when queried about how he got the list. "I just got it."

He then cross-checked bombing targets with media reports to identify locations for on-site investigation. Already he has begun to rule out incidents--either because of duplication or because the incidents didn't happen.

In Afghanistan, he's using a Global Positioning System to compare his ground location with the laptop map. The team examines bomb craters, destroyed structures and graveyards and interviews villagers as well as hospital and government officials.

On-site investigation is critical. "When you get into these things, you realize how often rumor differs from reality," says Human Rights Watch Executive Director Roth. In Yugoslavia, for instance, a report that a hospital was destroyed by bombing turned out to be a broken window. A report that bombs had knocked out the Bistrica hydroelectric plant turned out to be erroneous.

"You have to be there," he says. "There is no substitute for physical examination and talking to eyewitnesses."

###