How Likely Is 'Accidental' Nuclear War?

FIVE FALSE ALERTS of nuclear attack on the U.S. in the last two years have demonstrated that the United States' strategic early-warning system is unreliable.

The early warning system is designed to detect the launching of Russian missiles in a nuclear attack on this country. Signals from satellites and radars are fed into a network of computers focused at the North American Aerospace Defense Command Center (NORAD) in Colorado and three other command centers. The information analysed and displayed by the computers is constantly discussed by members of the four commands in "missile display conferences."

During a nuclear attack, Soviet missiles would arrive in at most 30 minutes after they have been detected, in ten minutes or less if launched from submarines. If the computer system seems to indicate the highly unlikely event of a completely unexpected attack, the U.S. military command quickly sends recallable bombers and control aircraft into the air, and a series of conferences is held. "Routine" missile display conferences are held whenever U.S. satellites or radars are changed because these changes can upset the computer system. Missile display conferences are also held whenever there is uncertainty about how to interpret the computer displays.

The next step in an alert, involving higher ranking personnel, is a "threat assessment conference." If the threat seems to be a real one, a "missile attack conference" may be called, bringing in the president and all senior Cabinet and 'White House personnel. There has never been a missile attack conference, and no detailed information about who would make decisions in such a situation is available to the public. The number of times the displays have had to be discussed is strikingly large. From January 1979 through June 1980 there were 3703 routine missile display conferences--an average of about six daily. Another 148 conferences, about two each week, were called because the displays were thought to indicate possible attacks. Five threat assessment conferences were also called:

October 3, 1979: A radar detected an orbiting rocket and generated a false launch and impact report (No public information on the action taken).

November 9, 1979: Data simulating a massive Soviet attack was accidentally fed into the NORAD computer system. Ten jet interceptors took off and the crews of about 1000 Minutemen silos were alerted.

March 15, 1980: A submarine-based missile launched during Soviet training generated an unusual signal. (No public information on the action taken).

June 3, 1980: A faulty circuit in a NORAD computer caused false messages to be sent to two command posts. Alert bomber crews were ordered to start engines, and the Pacific Command aircraft took off.

June 6, 1980: NORAD intentionally reproduced the situation in an attempt to track the fault. Similar events occurred.

It is extremely important for the United States not to launch a single nuclear weapon because of a false alert or--by making mistaken preparations--cause the U.S.S.R. to launch a real attack. And yet, the events cited above indicate that both of these scenarios are very real possibilities.

Sens. Gary Hart (D-Colo.) and Barry Goldwater (R-Ariz), both members of the U.S. Senate Committee on Armed Services, recently issued a report on the June 1980 alerts. Their report emphasized that it took only a few minutes to discover that there was no real attack, and that many steps must be taken before a missile can actually be launched. However, during the June 3 false alert the Pacific Command took off after the Commander of NORAD had confirmed that there was no threat. In the words of the report, "There seemed to be an air of confusion."

It is impossible to guarantee that future false alerts will be stopped so quickly. Tension and pressure that accompanies an alert would be heightened at a time of crisis between the U.S. and the U.S.S.R.

Effective control is also threatened by the unreliability of the computer communication links. In December 1968, NORAD devised a program for updating the computer equipment in the early warning system. This program "427M" has been repeatedly criticized by the General Accounting Office, the Air Force, and the Defense Communications Agency. Two main features were attacked:

Inadequate Equipment: The power supply is interruptible, much of the hardware is outdated, and the software is non-standard. This makes the early-warning system unreliable, slow and unable to distinguish the signals of a real launch from many other signals.

Split Management: The management of the system is inefficiently split between four parts of the Air Force. Cost over-runs already exceed $100 million and the system is not yet fully operational.

Computer equipment in the early-warning system and human reactions to alerts are both unreliable. If highly reliable communication links existed between the U.S. and the U.S.S.R., the danger of dependence on computer-warning systems would be reduced. But right now they don't. And strategies such as "launch on warning" or "launch under attack" only serve to increase the likelihood of an accidental nuclear war.

Sheena C. Phillips, a graduate of Cambridge University, England, is a researcher at the Union of Concerned Scientists.