"Circuit fully operational," reported Hal after only ten seconds. In that time, he carried out as many tests as a small army of human inspectors.

"Fine," said Poole with satisfaction. "Now replacing the cover."

This was often the most dangerous part of an extravehicular operation: when a job was finished and it was merely a matter of tidying up and getting back inside the ship – that was when the mistakes were made. But Frank Poole would not have been on this mission if he had not been careful and conscientious. He took his time, and though one of the locking nuts almost got away from him, he caught it before it had traveled more than a few feet.

Fifteen minutes later he was jetting back into the space-pod garage, quietly confident that here was one job that need not be done again.

In this, however, he was sadly mistaken.

"Do you mean to say," exclaimed Frank Poole, more surprised than annoyed, "that I did all that work for nothing?"

"Seems like it," answered Bowman. "The unit checks out perfectly. Even under two hundred percent overload, there's no fault prediction indicated."

The two men were standing in the tiny workshop-cum-lab in the carrousel, which was more convenient than the space-pod garage for minor repairs and exanimations. There was no danger, here, of meeting blobs of hot solder drifting down the breeze, or of completely losing small items of equipment that had decided to go into orbit. Such things could – and did – happen in the zero-gee environment of the pod bay.

The thin, card-sized plate of the AE-35 unit lay on the bench under a powerful magnifying lens. It was plugged into a standard connection frame, from which a neat bundle of multicolored wire led to an automatic test set, no bigger than an ordinary desk computer. To check any unit it was only necessary to connect it up, slip in the appropriate card from the "trouble-shooting" library, and press a button. Usually the exact location of the fault would be indicated on a small display screen, with recommendations for action.

"Try it yourself," said Bowman, in a somewhat frustrated voice. Poole turned the OVERLOAD SELECT switch to X-2 and jabbed the TEST button. At once, the screen flashed the notice: UNIT OK.

"I suppose we could go on turning up the juice until we burned the thing out," he said, "but that would prove nothing. What do you make of it?"

"Hal's internal fault predictor could have made a mistake."

"It's more likely that our test rig has slipped up. Anyway, better safe than sorry. It's just as well that we replaced the unit, if there's the slightest doubt."

Bowman unclipped the wafer of circuitry, and held it up to the light. The partly translucent material was veined with an intricate network of wiring and spotted with dimly visible microcomponents, so that it looked like some piece of abstract art.

"We can't take any chances – after all, this is our link with Earth. I'll file it as N/G and drop it in the junk store. Someone else can worry about it, when we get home."

But the worrying was to start long before that, with the next transmission from Earth.

"X-ray-Delta-One, this is Mission Control, reference our two-one-five-five. We appear to have a slight problem.

"Your report that there is nothing wrong with the Alpha Echo three five unit agrees with our diagnosis. The fault could lie in the associated antenna circuits, but if so that should be apparent from other tests.

"There is a third possibility, which may be more serious. Your computer may have made an error in predicting the fault. Both our own nine-triple-zeros agree in suggesting this, on the basis of their information. This is not necessarily cause for alarm, in view of the back-up systems we have, but we would like you to watch out for any further deviations from nominal performance. We have suspected several minor irregularities in the past few days, but none have been important enough for remedial action, and they have shown no obvious pattern from which we can draw any conclusions. We are running further tests with both our computers and will report as soon as the results are available. We repeat that there is no need for alarm; the worst that can happen is that we may have to disconnect your nine-triple-zero temporarily for program analysis, and hand over control to one of our computers. The time lag will introduce problems, but our feasibility studies indicate that Earth control is perfectly satisfactory at this stage of the mission.

"X-ray-Delta-One, this is Mission Control, two-one-five-six, transmission concluded."

Frank Poole, who was on watch when the message came in, thought this over in silence. He waited to see if there was any comment from Hal, but the computer did not attempt to challenge the implied accusation. Well, if Hal would not raise the subject, he did not propose to do so either.

It was almost time for the morning changeover, and normally he would wait until Bowman joined him on the control deck. But today he broke this routine, and made his way back to the carrousel.

Bowman was already up, pouring himself some coffee from the dispenser, when Poole greeted him with a rather worried "good morning." After all these months in space, they still thought in terms of the normal twenty-four-hour cycle – though they had long since forgotten the days of the week.

"Good morning," replied Bowman. "How's it going?" Poole helped himself to coffee. "Pretty well. Are you reasonably awake?"

"I'm fine. What's up?"

By this time, each knew at once when anything was amiss. The slightest interruption of the normal routine was a sign that had to be watched.

"Well," Poole answered slowly. "Mission Control has just dropped a small bomb on us." He lowered his voice, like a doctor discussing an illness in front of the patient. "We may have a slight case of hypochondria aboard."

Perhaps Bowman was not fully awake, after all; it took him several seconds to get the point. Then he said "Oh-I see. What else did they tell you?"

"That there was no cause for alarm. They said that twice, which rather spoiled the effect as far as I was concerned. And that they were considering a temporary switchover to Earth control while they ran a program analysis."

They both knew, of course, that Hal was hearing every word, but they could not help these polite circumlocutions. Hal was their colleague, and they did not wish to embarrass him. Yet at this stage it did not seem necessary to discuss the matter in private.

Bowman finished his breakfast in silence, while Poole toyed with the empty coffee container. They were both thinking furiously, but there was nothing more to say.

They could only wait for the next report from Mission Control – and wonder if Hal would bring up the subject himself. Whatever happened, the atmosphere aboard the ship had subtly altered. There was a sense of strain in the air – a feeling that, for the first time, something might be going wrong.

Discovery was no longer a happy ship.