This woman was paddling smoothly upstream, into the hissing water that only around her seemed to be re-collecting itself as a liquid. Upstream it was a steep rapids, leading to the white smash at the bottom of the falls proper.

The kayaker pushed up into this wilder section, paddling harder upstream, then held her position against the flow while she studied the falls ahead. Then she took off hard, attacking a white smooth flow in the lowest section, a kind of ramp through the smash, up to a terrace in the whitewater. When she reached the little flat she could rest again, in another slightly more strenuous maintenance paddle, gathering her strength for the next salmonlike climb.

Abruptly leaving the strange refuge of that flat spot, she attacked another ramp that led up to a bigger plateau of flat black water, a pool that had an eddy in it, apparently, rolling backward and allowing her to rest in place. There was no room there to gain any speed for another leap up, so that she appeared to be stuck; but maybe she was only studying her way, or waiting for a moment of reduced flow, because all of a sudden she attacked the water with a fierce flurry of paddle strokes, and seemingly willed her craft up the next pouring ramp. Five or seven desperate seconds later she leveled out again, on a tiny little bench of a refuge that did not have a pushback eddy, judging by the intensity of her maintenance paddling there. After only a few seconds she had to try a ramp to her right or get pushed back off her perch, and so she took off and fought upstream, fists moving fast as a boxer’s, the kayak at an impossible angle, looking like a miracle until all of a sudden it was swept back down, and she had to make a quick turn and then take a wild ride, bouncing down the falls by a different and steeper route than the one she had ascended, losing in a few swift seconds the height that she had taken a minute or two’s hard labor to gain.

“Wow,” Frank said, smitten.

She was already almost down to the hissing tapestry of flat river right below him, and he felt an urge to wave to her, or stand and applaud. He restrained himself, not wanting to impose upon another athlete obviously deep in her own space. But he did whip out his cell phone and try out a GPS-oriented directory search, figuring that if she had a cell phone with a transponder in the kayak, it had to be very close to his own phone’s position. He checked his position, entered thirty meters north of that; got nothing. Same with the position twenty meters farther east.

“Ah well,” he said, and stood to go. It was sunset now, and the smooth stretches of the river had turned a pale orange. Time to go home and try to fall asleep.

“In search of kayaker gal, seen going upstream at Great Falls. Great ride, I love you, please respond.”

He would not send that in to the free papers, but only spoke it as a kind of prayer to the sunset. Down below the kayaker was turning to start upstream again.

IT COULD be said that science is boring, or even that science wants to be boring, in that it wants to be beyond all dispute. It wants to understand the phenomena of the world in ways that everyone can agree on and share; it wants to make assertions from a position that is not any particular subject’s position, assertions that if tested for accuracy by any sentient being would cause that being to agree with the assertion. Complete agreement; the world put under a description stated that way, it begins to sound interesting.

And indeed it is. Nothing human is boring. Nevertheless, the minute details of the everyday grind involved in any particular bit of scientific practice can be tedious even to the practitioners. A lot of it, as with most work in this world, involves wasted time, false leads, dead ends, faulty equipment, dubious techniques, bad data, and a huge amount of detail work. Only when it is written up in a paper does it tell a tale of things going right, step-by-step, in meticulous and replicable detail, like a proof in Euclid. That stage is a highly artificial result of a long process of grinding.

In the case of Leo and his lab, and the matter of the new targeted nonviral delivery system from Maryland, several hundred hours of human labor and many more of computer time were devoted to an attempted repetition of an experiment described in the crucial paper, “In Vivo Insertion of cDNA 1568rr into CBA/H, BALB/c, and C57BL/6 Mice.”

At the end of this process, Leo had confirmed the theory he had formulated the very moment he had read the paper describing the experiment.

“It’s a goddamned artifact.”

Marta and Brian sat there staring at the printouts. Marta had killed a couple hundred of the Jackson labs’ finest mice in the course of confirming this theory of Leo’s, and now she was looking more murderous than ever. You didn’t want to mess with Marta on the days when she had to sacrifice some mice, nor even talk to her.

Brian sighed.

Leo said, “It only works if you pump the mice full of the stuff til they just about explode. I mean look at them. They look like hamsters. Or guinea pigs. Their little eyes are about to pop out of their heads.”

“No wonder,” Brian said. “There’s only two milliliters of blood in a mouse, and we’re injecting them with one.”

Leo shook his head. “How the hell did they get away with that?”

“The CBAs are kind of round and furry.”

“What are you saying, they’re bred to hide artifacts?”

“No.”

“It’s an artifact!”

“Well, it’s useless, anyway.”

An artifact was what they called an experimental result that was specific to the methodology of the experiment, but not illustrating anything beyond that. A kind of accident or false result, and in a few celebrated cases, part of a deliberate hoax.

So Brian was trying to be careful using the word. It was possible that it was no worse than a real result that happened to be generated in a way that made it useless for their particular purposes. Trying to turn things that people have learned about biological processes into medicines led to that sort of result. It happened all the time, and all those experimental results were not necessarily artifacts. They just weren’t useful facts.

Not yet, anyway. That’s why there were so many experiments, and so many stages to the human trials that had to be so carefully conducted; so many double blind studies, held with as many patients as possible, to get good statistical data. Hundreds of Swedish nurses, all with the same habits, studied for half a century but these kinds of powerful long-term studies were very rarely possible. Never, when the substances being tested were brand-new literally, in the sense that they were still under patent and had brand names different from their scientific appellations.

So all the little baby biotechs, and all the start-up pharmaceuticals, paid for the best stage-one studies they could afford. They scoured the literature, and ran experiments on computers and lab samples, and then on mice or other lab animals, hunting for data that could be put through a reliable analysis that would tell them something about how a potential new medicine worked in people. Then the human trials.

It was usually a matter of two to ten years of work, costing anywhere up to five hundred million dollars, though naturally cheaper was better. Longer and more expensive than that, and the new drug or method would almost certainly be abandoned; the money would run out, and the scientists involved would by necessity move on to something else.

In this case, however, where Leo was dealing with a method that Derek Gaspar had bought for fifty-one million dollars, there could be no stage-one human trials. They would be impossible. “No one’s gonna let themselves be blown up like a balloon! Blown up like a goddamn bike tire! Your kidneys would get swamped or some kind of edema would kill you.”