A balky backup computer today forced the maiden flight of the space shuttle Columbia to be postponed, until Sunday at the earliest.

Launch Control Director George F. Page scrubbed the flight of astronauts John W. Young and Robert L. Crippen at 9:59 a.m., when flight controllers could not get the backup computer to respond to repeated commands that it communicate with the shuttle's four prime computers.

The postponement came more than three hours after Young and Crippen were scheduled to take Columbia on a 54-hour test flight 170 miles above the Earth. It was the first time since 1966 that a space flight has been scrubbed with astronauts on the launch pad.

Flight engineers isolated the computer problem tonight and were working to prevent it from happening again, but the earliest Young and Crippen can fly in any event will be Sunday. It takes that long to recondition the shuttle's external fuel tank.

The 1.5 million gallons of super-cold liquid hydrogen and oxygen must be emptied into insulated underground tanks while the rest of the countdown process is under way. Otherwise, it would evaporate, in effect boiling away, because it is so cold. It then has to be piped back into the external tank. That whole process takes a little more than a day.

"Any time we scrub on the pad you're looking at a 48-hour delay," said Clyde Netherton, chairman of the shuttle countdown working group. "It takes at least a day to get that hydrogen tank back in condition to accept fuel."

The three major television networks tentatively have rescheduled live coverage of the launch, which is again set for 6:50 a.m. ABC said it would cover the event from 6 a.m. to 8 a.m., CBS indicated it would start coverage at 6:30 or 7:30 a.m., and NBC said its coverage could start some time between 5:30 and 7:30 p.m.

The countdown had proceeded almost without a hitch. Young and Crippen had been in Columbia's cockpit today for two hours when a red light flashed and a message came across the display screen in front of them.

The backup computer, designed to take over the flight controls on ascent and re-entry if the four main computers fail, was not "talking" as it should with two of the other four computers.

"We all saw it even though it was something we'd never seen before," said Neil B. Hutchinson, ascent flight director at Houston's Johnson Space Center. "There's always a chance you'll run up against something you've never seen before."

Through thousands of hours of tests, flight controllers had seen the backup computer miss communicating on occasion with one of the other four computers, but never two. Also, controllers always had been able to correct the mistake by pushing a reset button that controls the computer's traffic flow. This time, that didn't work.

"The crew did a reset, we did a reset here in Houston, they tried again and we tried again," Hutchinson said. "We ran out of our bag of tricks."

If the backup computer fails to communicate with all four of Columbia's on-line computers at launch, there is a chance that the flight could turn into a catastrophe.

The four computers fly the shuttle all the way into orbit, navigating, throttling the engines and commanding the flow of fuel to the engines. Should the main computers suffer a sudden generic failure -- all failing for the same underlying reason -- the backup computer is programmed to take over the flight at once.

"The backup kind of goes to sleep in orbit, but on ascent and entry it's right up to speed, ready to take over if all four fail," Hutchinson said. "We've got to have a working backup at liftoff and re-entry. We never take off without all five."

Hutchinson explained that if the backup had failed before re-entry, the crew could have replaced it with a sixth, spare computer it carries in flight, a procedure that takes about two hours.

But during the countdown this morning there wasn't enough time to replace the backup computer after they learned which one was malfunctioning. Flight officials didn't know whether the trouble was in the backup, the four main computers, the backup's program or the programs of the other four.

"We've already dumped all the data that's stored on the backup and all four mains and are examining it to see what could have gone wrong," said Brock R. Stone, Data Processing Systems Engineer at Johnson Space Center.

Between 50 and 100 computer engineers at the Kennedy Space Center, the Johnson Space Center, IBM in Oswego, N.Y., and Rockwell International in Downey, Calif., where the computers are built and programmed, began trouble-shooting, trying to duplicate what had gone wrong this morning to figure it out and then fix it.

By 9 tonight, engineers had figured that the problem was caused by the software in the four prime computers, which were "out of sync" with the backup computer. The prime computers were fractions of a second off in telling the backup computer what they were doing, which was interpreted by the backup computer as not receiving any information.

Engineers still had no quick way to solve this problem, but felt they were well on their way to fixing it. If they don't succeed, there is a good chance that the launch will be postponed again.

Hutchinson and Stone made it clear today that the launch would not be rescheduled until flight engineers fully understand what caused the backup computer to turn so erratic. Mission managers were to hold a telephone conference meeting by mid-afternoon Saturday to decide whether they were ready to launch Sunday.

There was no word from astronauts Young and Crippen about how they felt about the postponement today of the flight they have trained four years to make. Hutchinson was asked their reaction to the scrub, and he replied: "They took it in stride. John Young said, 'Well, you all did real good. I'm just sorry we couldn't go.'"