The jagged blast of a siren shook the control room. Warning lights began to blink. The computer started printing question marks. Almost immediately, the operators of the Three Mile Island power plant realized that they were facing a nuclear accident that neither man nor machine had been programmed to handle.

"It was sufficiently abnormal," said Ed Frederick, one of the five control room operators on duty March 28 at the time of the Pennsylvania nuclear accident, "to be just completely out of my realm of experience."

"I had not been in this situation, and did not ever imagine that I would be," said Frederick's supervisor, Gary Miller. ". . . The operator did not have enough information."

A House task force investigating Three Mile Island yesterday released transcripts of its interviews with Frederick, Miller, and their coworkers about the situation in the control room during the minutes the nation's worst nuclear accident was starting.

The transcripts marked the first public presentation of the control room operator's description of what happened. Senior officials of the Nuclear Regulatory Commission and the utility that owns Three Mile Island have discussed the accident at length, but the operators have not yet been heard outside closed hearing rooms.

The transcripts give a vivid picture of confusion and frustration as the operators tried to deal with a situation that was, according to all their training, impossible.

Frederick and another operator, Craig Faust, said they were uncertain how to react when pumps on two safety systems failed simultaneously because such an accident simply could not happen on the "simulator" where they had learned how to run the plant.

Like the operators of other nuclear plants designed by Three Mile Island's builder, Babcock & Wilcox Inc., Frederick and Faust had been trained on a Babcock & Wilcox simulator-a mock control room connected to a computer that mimics reactor conditions.

Babcock & Wilcox officials told the task force that they have subsequently changed the simulator program to include the type of mishap that occurred in Pennsylvania.

Frederick, however, said the simulator's shortcomings were not the most serious fault of the plant operators' training program.

The training program, Frederick said, is designed to get potential operators past the Nuclear Regulatory Commission's licensing examination.

"They know the type of questions the NRC is going [to] ask, the type of response that they want to emergency procedure questions . . . you are supposed to memorize them and reproduce them word for word on a piece of paper.

"That type of examination procedure, I think, is detrimental to expanding your training, in other words," Frederick said.

Officials at Babcock & Wilcox and the NRC have said that mistakes by the control room operators in the first minutes after the accident exacerbated the harm to the reactor and made the incident more serious than it might have been.

In their interviews with the House task force, the operators tried hard to counter that suggestion. They said thhey were misled by faulty instrumentation and by emergency procedures that seemed to fly in the face of common-sense reactions to their problems. And the operators returned repeatedly to flaws in their training.

"I think the training philosophy was inadequate," Frederick said. "The basic assumption is that the design encompasses enough emergencies and anticipated casualties, that what happened there [at Three Miles Island] probably couldn't happen."

This prompted the task force chairman, Rep. Jim Weaver (D-Ore.), to think of chess. "You know," he said, "there are 2 billion possible moves in the first four moves of chess-2 billion.

"Is that what we've dealing with here-an infinite number of things that could go wrong?"