It promised to be a quiet evening at the Soviet nuclear early warning center when Lt. Col. Stanislav Petrov settled into the commander's seat on Sept. 26, 1983.
But within minutes, Colonel Petrov was locked in perhaps the most dangerous drama of the cold war. An alarm sounded, warning screens blinked. A computer map on the wall showed the hostile launch of a US nuclear warhead.
"Every second counted.... My legs were unsteady, my hands were trembling, my cozy armchair became a hot frying pan," says the former officer. It only got worse. Within five minutes the computer registered five more launches; the alarm flashed: "Missile Attack."
The decision that Petrov made in those pressure-cooked minutes - that the computer was in error, and the elaborate early warning system that he helped build was wrong - may have prevented a nuclear holocaust.
Twenty years later, there is growing concern that a similar nuclear miscue could happen again. The lone superpower and its former rival still aim thousands of missiles on hair-trigger alert at each other's major cities. As the US rushes to deploy a missile shield this summer designed to intercept North Korean warheads, Clinton-era plans that would improve both US and deteriorating Russian detection systems are stalled.
At presidential summits in both 1998 and 2000, the US and Russia announced plans for a joint, real-time warning system in Moscow. The blueprint, drawing on American's sophisticated satellite network and Russia's wide radar net, promised to keep better tabs on the superpower arsenals as well as on terrorist threats.
"I wish I could say there is no chance of it [today]," Petrov says, in a matchbox kitchen with a yellowing star chart. "But when we deal with space - when we [play] God - who knows what will be the next surprise?"
Dreams of joint efforts ground to a halt long ago. Meanwhile, neglect has left Russia's system in disrepair.
"The fact is, the Russians are flying blind," says Jon Wolfsthal, a nonproliferation expert at the Carnegie Endowment for International Peace in Washington. "There are huge portions of their periphery that are unmonitored because their satellites are down, and they've lost a number of [Soviet-era] radar sites."
The result is a growing concern about false readings that could show a hostile nuclear launch, and provoke real retaliation. Such fears have been augmented by a string of Russian military accidents, from failed test missile launches to a sunken nuclear submarine.
After a secret year-long investigation into the 1983 incident, Petrov says the false readings that shocked him and his team were attributed to a rare but predictable reflection off the earth. The system was fooled again in 1995, when Russians briefly thought that a scientific launch from Norway was a nuclear-tipped US missile heading their way. President Boris Yeltsin reportedly brought out the launch suitcase called the "nuclear football" - perhaps the closest it's ever come to being used in Soviet or Russian history - before coming to believe there was no need to respond.
"There are examples of weather satellite launches, the full moon rising, flocks of geese - all these horror stories in history," says Mr. Wolfsthal. Part of the solution was meant to be a joint warning center built in Moscow. It was to have been completed years ago.
When President Bill Clinton and Mr. Yeltsin first announced plans to build the center in 1998, it was heralded as a breakthrough in preventing a "false warning" leading to accidental war. When Russia's new President Vladimir Putin signed the deal in 2000 with Mr. Clinton, the White House touted it as a "milestone in ensuring strategic stability."
Expert advice backed up that view, with compelling findings in the mid-1990s that combining the US and Russian systems would significantly boost results.
"We went through a whole simulation of different missiles fired by different countries from the Mideast to Europe at different targets at different trajectories," says Bruce Blair, president of the Center for Defense Information in Washington who oversaw the detailed research with Russian scientists.
"We looked at detections of the US system operating alone, and the Russian one alone, and found the combined performance would be 20 to 70 percent better," says Mr. Blair, a former Minuteman launch officer and early proponent of the joint warning center.
With seven Russian radars stretched from the Baltics to Azerbaijan - which can pick-up Middle East launches - and the US satellites, "you just have a lot more assets focusing on the threat," adds Blair. "We could work together and have a much better ability to detect third country attacks."
US and Russian leaders had another reason to be optimistic. Russian missile officers used a US command center in Colorado Springs, Colo., for months at the end of 1999, to familiarize themselves with the US early-warning system - and to be on hand during the Millenium New Year - to ensure direct contact with Moscow in case any Y2K computer "bug" affected the Russian system.
"It was incredibly useful, and built a lot of trust between ... early-warning groups on both sides," says Carnegie's Wolfsthal. "But in the end, they went away, and we're left without real-time sharing."
The joint project, first envisioned for completion in mid-2001, has foundered on everyday issues of what Russian taxes should be paid for imported US equipment, and legal concerns about liability.
"It's a lack of political will on both sides," says Vladimir Dvorkin, a former major general in Russia's nuclear forces, and a top strategist until 2001. Russia's early warning is "less trustworthy" than the American one, he says, though its readings "will never serve as the only guide [to launch]."
Even more important, General Dvorkin says, is "joint analysis of threats coming from unstable regions with authoritarian regimes."
The mundane points stalling the project are surprising security experts. "If you're a lawyer at the State Department, [liability and taxes] may be very important issues," says Wolfsthal. "But if you are concerned about the geostrategic survival of the human species, they are minuscule in their relevance."
Experts on both sides are planning to issue a wake-up call, however. A group has planned to begin a year-long modeling exercise that will examine the risks of inaction, and the need for the joint center.
"We want to show what can happen without this center," says Pavel Zolotarev, a former Strategic Forces major general. "If such a center were already open here, these threats would probably not lead to dangerous situations. We'll get these results in a year, but who knows when we will be able to convince the leadership?"
Over the past decade, the US has spent roughly $7 billion funding nuclear-threat-reduction programs to control "loose nukes" and to secure weapons-grade nuclear material and scientific expertise that might be easy targets for terrorists. The $1 billion total spent per year on all threat reduction amounts to less than one-third of one percent of US defense spending.
Taking the thousands of warheads off hair-trigger alert, experts say, would also help lower risks. Experts estimate that there is a world total of 30,000 assembled nuclear weapons and enough bomb-grade material to create nearly a quarter million more.
Some Russian experts argue that, since they have no intention of ever "launching on warning," Moscow is deliberately letting its early-warning system deteriorate.
"Certainly if you look at what they do to maintain their satellite constellation, it's pathetic," says Matthew Bunn, a nuclear expert at Harvard's Belfer Center. There is virtually no coverage of areas where US submarines could launch missiles - long considered the first phase of any attack on Soviet forces.
"I can't imagine a Russian president awakened in the middle of the night with a blip on the radar screen, saying 'Yes, I'm going to launch the missiles on that basis,' knowing what he has to know about the state of their early-warning system," says Mr. Bunn.
• Second in an occasional series on US-Russian strategic issues. "Old weapons, new terror worries," ran April 15, 2004.