“Talk to me, Donald. Give me an evidence summary.”

“Yes, sir. Several new facts have come to light from the hospital and our forensic lab. First and foremost, we have confirmed that the bloody footprints match the tread patterns of a standardized robot body model manufactured at Leving Labs. That robot body is a large general-purpose model, used with various brain types and body modifications for various purposes. The length of the footprints’ stride precisely matches that of the standard specification for that robot body model. The wound on Fredda Leving, s head corresponds to the shape and size of the arm of the same robot type, striking from the rear and to the left of the victim, from an angle consistent with Fredda Leving, s height and the height of that robot model-though all of those measurements are approximate, and any number of other blunt instruments would match, and a whole range of heights, forces, and angles would also be consistent with the wound.

“Microtraces of a red paint found in Madame Leving’ s scalp wound likewise correspond to a paint used on some robots at Leving Labs, though it has not been definitively established that the paint in question was used on the robot model in question. I might add that it could not be immediately established whether the microtraces were from fresh or fully dried and hardened paint, as it was some hours before the labtech robots secured the samples. Further tests should answer that question.”

“So the only suspect we are offered is a robot. That’s impossible, of course. So it had to be a human-a Settler-posing as a robot. Except even a Settler who had been on the planet five minutes would know that it is impossible for a robot to attack a human. Why bother to plant doctored evidence we will refuse to believe?”

“That point has bothered me as well,” Donald said. “But even if we assume a Settler was involved in this crime, we must assume that the Settler in question knew more about robots than the average Spacer.”

“What do you mean?”

“Consider the detailed familiarity and access to robot equipment required to stage this attack,” Donald replied. “The assailant would have to build and wear shoes with robot foot-tread soles and then replicate the gait of a specific robot. He or she would have to use a surplus robot arm-or an object that closely matched it-as a blunt instrument, and strike in such a manner as to match a blow from that robot arm. He or she would need access to the proper materials to stage the attack, and have the mechanical skill to build or modify the needed robot body parts. To be blunt, sir, a human capable of staging this attack could not possibly be stupid enough or sufficiently ignorant of robots to dream that we would think a robot did it.”

“But then what was the motive for staging the attack in that manner?” Kresh asked. He thought for a moment. “You said this footprint and arm are off a very standard robot model. How many of them out there?”

“Several hundred. Several thousand if you include all the variants.”

“Very well, then. That means there have been several thousand opportunities to steal a robot, or secure a defective one, and strip it for parts-the feet and arms and so forth. Or hell, the assailant could simply get hold of a robot and yank the positronic brain. He or she could plug in a remote-control system with a video-link back to the controller. Let the controller walk the robot body up to the unsuspecting victim-after all, who would suspect a robot?

“And using a remotely operated robot body that would look like a normal robot would have to be less suspicious than wearing robot-tread boots and carrying a robot arm around. And by working from a remote location, the assailant could hide his or her identity. Another thing: If I popped someone on the head, I’d want to get awayfast. Yet those footprints were of a walking gait, not a run. That points toward a remotely controlled robot, one with a fairly limited remote-control system that could manage a walk but not a run.”

“Except the attacker did not leave immediately. He or she-or it-remained for some time after the attack, at least thirty seconds or a minute.”

“How do you know that?” Kresh asked. “ Ah, of course, the footprints. They went through the outer edges of the pooled blood, so they had to have been made after Leving had bled long enough to produce a large pool of blood. Damn it! That makes no sense. Why the devil would the attacker stay behind? Not to make sure Leving was dead, obviously, because she wasn’t. But we’re digressing. You suggested that the assailant would know that we’d know that a robot could not commit the crime. Therefore, the assailant had an alternate motive for disguising the attack as coming from a robot. What would that be? Why such an elaborate setup?”

“To afford the chance to get lost in the crowd later,” Donald suggested. “Let me offer a hypothetical variation on the facts by way of example. We now have an impossible suspect, a robot. Let me offer another impossible suspect to make my point, though I must ask you not to take offense at a hypothetical example.”

“Of course not, Donald. Go on.”

“All right. If someone decided to plant clues to make it appear that, for example,you had attacked Fredda Leving, that would limit the search for the assailant to those persons with the ability to plant those clues. Someone who could steal a pair of your shoes, or manage to plant strands of your hair, or your fingerprints, at the scene. But if that someone chose to plant clues that pointed equally well to several thousandidentical and impossible suspects-”

“Our search is made far larger. Yes. Yes, I see that. An excellent point, Donald. But there is still another question. What of the second set of footprints?”

“If you will grant, for the sake of argument, my original premise, that the effort to make this seem like a robot attack was made because we would know it was impossible a robot did it, I can offer an answer. If we further assume that the motive for that nonconvincing subterfuge was to disguise the real assailant, then I suggest that a single assailant deliberately made one set of bloody tracks, walked far enough that all traces of the blood were worn off, then simply doubled back and walked through the blood again. Again, the idea would be to confuse the search.”

“It seems an awfully risky thing to do for a fairly minor advantage,” Kresh objected.

“If, as you suggest, the attacker was using a remotely controlled robot body, as opposed to merely wearing robot boots and carrying a robot arm, there could be no risk in the gambit. At worst, someone might have come in during the assailant’s absence and been there to capture the false robot, with the real attacker at the controls, perhaps many kilometers away.”

“Yes. Yes. Now they would have us looking fortwo robots, ortwo people trying to disguise themselves as robots, when there was really only a single, human assailant. That’s a lovely theory, Donald, just lovely.”

“There is another note: Our robopsychologists have completed the preliminary interrogation of the staff robots at Leving Labs. Their results are, I think, astonishing.”

“Are they indeed?” Kresh asked dryly. “Very well, then, astonish me.”

“First, this is by no means the first time the staff robots have been instructed to stay out of the main wing of the labs. They have been told to get out many times before, usually but not always at about the hour of the attack, but always when the lab was more or less empty. This merely confirms what Daabor 5132 told me the night of the attack. However, the second point provides fresh and remarkable data.”

“Very well, go on.”

“Every single robot flatly refused to identify who had given the order. Our robopsychologists unanimously agree that the block restraining them is unbreakable. The psychologists took several robots to and past the breaking point, pressuring them to answer, and all refused to talk right up to the moment they brainlocked. The robots died rather than talk, even when told that their silence might well allow Fredda Leving’s attacker to go free.”