The survey by the New York-based International Rescue Committee (IRC) was conducted only in the rebel-held eastern half of the country, where most of the fighting and even more of the accompanying hardship has taken place. The toll would almost surely be higher if the government-held western half were included.

The survey attributes a relatively small proportion of the deaths -- a few hundred thousand -- to the battles waged by the Congolese army, its rebel foes and troops from the half-dozen other African countries that have fought on both sides of the conflict. The vast majority of deaths have resulted from starvation, disease and deprivation on a scale emerging only as aid workers reach areas that have been cut off by fighting and lack of roads.

"People are dying of nothing, of everything," said a worker for Pharmacists Without Borders, after returning from a humanitarian assessment in the Kasai region. Villagers in the region -- long renowned for its diamond mines, but now ravaged by hunger -- refer to two kinds of gems: white ones and red ones. The red ones are peanuts.

The IRC's figures are one-third higher than the number of deaths estimated from 18 years of war in Sudan, and three times the most frequently quoted death count for the Biafra conflict of the 1960s.

"Things are a little worse than the picture we painted last year," said Les Roberts, an epidemiologist formerly with the Centers for Disease Control and Prevention who conducted this year's IRC survey and one last year that estimated the death toll in eastern Congo at 1.7 million. "And last year's estimate turns out to have been low."

Until the IRC sent survey teams into eastern Congo, little was known about the human toll of a war fought largely out of the world's sight.

The conflict began in August 1998, when rebels backed by Rwanda and Uganda rose up against Laurent Kabila, then the president of Congo. After Angola, Zimbabwe and Namibia sent troops, warplanes and armor to aid Kabila's army, the war settled into a stalemate that has left the vast country divided roughly in half.

An agreement to bring in U.N. peacekeeping troops has moved forward since Kabila's assassination in January put his son, Joseph, in power. The new President Kabila frequently quotes IRC's first death estimate, calling it evidence that "this stupid war" verges on genocide.

Independent experts who have reviewed both IRC reports say the surveys appear to be sound. Epidemiologists offer two explanations for the extreme numbers: Strife in Africa usually goes unmeasured, and the strife in Congo has gone on for almost three years, perhaps a dozen times longer than the epidemics and carnage that typically produce the stunning death rates that IRC reports.

"It may be that this situation is worse than any of these others," said Ronald Waldman, a specialist in refugee health at the Mailman School of Public Health at Columbia University. "This is a more chronic, deeply cutting episode, I think."

Added another Western medical epidemiologist with long experience in humanitarian emergencies: "My personal belief is these numbers are the absolute best that could be done in the circumstances, and there's absolutely no reason to believe any bias of any kind has found its way in. The mortality is so outrageously high that even if the numbers were half that, the mortality is still outrageously high."

Although a National Academy of Sciences committee approved the methodology of the original study, IRC concedes that there were structural imperfections in both tallies. Roberts emphasized the risks inherent in extrapolating a death rate for a whole province based on house-to-house interviews in sites that surveyors believe -- but, without solid population data, cannot know -- to be representative.

Yet some of IRC's key findings -- of extremely high mortality among adults and extraordinary death rates among young children -- were replicated by a second aid agency in a January survey of one relatively peaceful district of eastern Congo. Around Kalima in Maniema province, the British medical aid group Merlin documented 2 1/2 times more deaths than births in a population that before the war had been growing by 3 percent a year. The group found the result so alarming it urged further investigation "to ensure these figures are accurate."

The estimates for Congo may be higher than those for other African conflicts partly because of the country's size. Congo's population is estimated at roughly 50 million -- of whom one-third are vulnerable to starvation and 2 million have been driven from their homes by fighting, according to the United Nations.

While the various studies point to similar conclusions, it is the IRC findings that beggar belief even among some war zone demographers.

"One doesn't know what to do with that kind of estimate except reach down and pull your jaw up off the floor," said Jeff Drumtra, a researcher for the U.S. Committee for Refugees, which calculated the death toll in Sudan from a review of news accounts, aid agency reports and population studies. "I just feel that if it's 1.7 million -- or even now above that -- either [Roberts is] terribly wrong or we've all been wrong about the death tolls in every African war. Ever."

Roberts said the surveys would stand up to scrutiny, as the first report apparently did in briefings to U.S. officials and aid agencies last summer. The basic approach is a time-tested method of measuring suffering in conflicts.

What makes the Congo surveys exceptional is not only their relative ambition, but how long the conditions they document have been allowed to persist.

"Mortality rates this high are common in humanitarian emergencies," said the Western epidemiologist, who identified similar death rates in Liberia, Somalia and a handful of other African hot spots. "But they only last a couple of months and then they lapse into some sort of background rate. Because there's some sort of intervention."

In Congo, the hugely elevated mortality rates have continued for 32 months, steadily racking up deaths by the hundreds of thousands across a vast region rendered inaccessible to aid because of fighting and the lack of roads.

For its latest estimate, IRC ventured into eight health districts scattered in the five provinces that historically define eastern Congo, an area of 20 million inhabitants. With the help of IRC country director Michael Despines, Roberts selected roughly representative zones that could be reached by plane or secure roads. Asked whether concerns about safety might bias the sample, Roberts said they surely did -- against finding the worst. Last year's IRC survey showed the most remote areas were the deadliest.

"I am sure there's a self-selection," said Roberts, who grew out his beard to encourage the protective assumption, common at military checkpoints, that a white man in Congo must be a priest, "and the self-selection bias is I'm a freakin' coward!"

Once a district is chosen, however, bias should be thwarted by technology. A hand-held global positioning device generates coordinates that direct survey-takers to particular houses, leaving no option for being drawn subconsciously to, say, the doorstep of a malnourished child. IRC sent trained African staff members to 1,460 homes, where inhabitants were asked: How many people live here? And have any died in the last year?

The results were tallied and compared with the number of deaths that the Centers for Disease Control and Prevention say should be expected in a country as decrepit as Congo: 1.5 deaths per 1,000 people per month. The rate assumes that, in peacetime, Congolese die at three times the rate of Americans. The IRC teams found the war has multiplied that another two to eight times.

Thus total deaths in Congo's five eastern provinces from January 1999 to May 2000 were calculated at 2.3 million -- 1.7 million more than what would have been expected in the absence of war. This year's survey, though still being calculated, was on track to add approximately a million more.

Children appeared to be perishing at an extraordinary rate. Around Moba, on the western shore of Lake Tanganyika, last year's survey showed nearly half of the infants were dying before reaching their first birthday. The figure was so staggering that Roberts scaled it back when extrapolating the data for the entire province, Katanga.

But last month, when Roberts brought a team back to Katanga, this time concentrating on the district of Kalemie, he found the situation was worse: Three of four children were dying before age 2.

"It's damn rare we see these kinds of hardship," Roberts said, using famines as reference points. "Probably in Somalia in the 1990s when we saw these skinny kids on TV. Probably in Ethiopia in 1984 you saw numbers like that."

IRC estimates that 40 percent of Congo's wartime deaths could have been avoided by access to the kind of basic health care that was becoming scarce even before the rebellions that have defined Congo over the last five years (an earlier rebellion took place from 1996 to 1997). At the start of the current war, only two of 11 provincial health inspectors had working vehicles.

Chiza Chilimu, 23, survived on fruits and wild yams for a year in the bush after Congolese militiamen called Mai-Mai emptied his village, Cibinda. But his mother died of relentless diarrhea.

"She died because she didn't have medicine," Chilimu said. "If we were at home, I could look for medicine."

Further aggravating the situation is Congo's economy, the world's fastest-shrinking since the war began. Said Roberts: "What's the main killer of the war? The main killer is people who don't have enough money in their pocket so they can take their kid to the clinic in time."

Even so, IRC estimated that 200,000 of the first 1.7 million deaths were by violence, almost all of them civilians. Even as a cease-fire has held on the front lines, massacres have continued behind rebel lines where militias and rebel armies terrorize some of Congo's more densely populated districts, attacking villages on the suspicion inhabitants are helping the other side.

Some groups are simply homicidal. The men with wild hair and bushy beards who kicked in the door of Jean Pierre Mushobozi's hut in the village of Buguli one night in November were Hutu extremists from Rwanda, who are associated with the group that carried out the 1994 genocide that killed more than a half-million Rwandan Tutsis and are now fighting for the Congolese government.

They slashed to death the two youngest children, slashed the leg of another and carried away the family's goats, hens, money and clothes -- even the spoons.

Congo's vast jungles are now populated by untold thousands who no longer feel safe in their homes but are ill equipped for living in the bush.

"People hide in the forest," said Claude Jibidar, head of the World Food Program office in Bukavu, capital of South Kivu province, where 380,000 people have been driven from their homes, many into the forest. "You don't really see them. You don't see the bodies."

World Food Program workers in South Kivu are returning from newly accessible villages with reports of more malnourished adults than children. "How come? Simply because the children are dead," Jibidar said.

"Basically, at some point malnutrition will be reduced," he said. "But it will be because the weakest have died."

As assessments of Congo's devastation accumulate, help has been very slow in coming. A January plea from the World Food Program to more than double its Congo food aid to $110 million has been barely one-third funded by rich countries. UNICEF has received just a tenth of the $15 million needed for essential drugs and therapeutic feeding centers. And despite vows of action from Washington that greeted the IRC's first survey, U.S. disaster relief to Congo remains at just $13 million. Halfway through the fiscal year, that sum is already exhausted.

The growing prospects for peace mean that "for the first time, people are optimistic, even myself," said Kees Tuinenburg, head of the World Food Program in Congo. "Has this translated into more humanitarian assistance? I don't think so."

Waldman, the Columbia professor, said the inattention is one reason the problem has grown to proportions that challenge belief.

"Looking back, we may have been on this slope for some time," he said. "Africa has been neglected for a long time, politically, economically, in just about every way possible. What did we expect? Last year is the first time any serious attention was given to . . . AIDS there. The epidemic has been going on for 20 years."