And Badim clearly wanted to mollify his old friend. “All right, point taken. Maybe we forgot too much, or took too much for granted.”

Freya said, “Now we’re not going to be facing any choices as stark as that, I hope. We’re on our way back to Earth, and there’s very little to do, given that project, but to keep things going well. Pass the ship along to our children in good working order, and teach them what they have to know. That’s what our parents did for us, as best they could. So now we do the same, and a few more generations do the same, and the last one in the line will be back on the planet we were made for.”

So they reestablished the general assembly, this time as including everyone aboard, all voting on issues that a working executive council deemed important. Voting was mandatory. The executive council was formed of fifty adults, drawn by lot for a five-year term, with very few acceptable reasons for getting off the council if one’s name were drawn.

Maintenance of the ship was left to us, with reports to the executive council and recommendations for human action included. We agreed to perform these functions.

“Happy to do so,” we said.

Literally? Was this a true feeling, or just an assertion? Could humans make that distinction when they said such things?

Possibly a feeling is a complex algorithmic output. Or a superposed state before its wave function collapses. Or a collation of data from various sensors. Or some kind of total somatic response, an affect state that is a kind of sum over histories. Who knows. No one knows.

The first new generation passed their second birthdays, and most of them began to walk a little before or after that time. It took a few months more to be sure that as a cohort their ability to walk was coming much later than had been true for earlier generations in the ship. We did not share this finding. However, as it became more statistically significant it also became more anecdotally obvious, and soon became one of those class of anecdotes that got discussed.

“What’s causing this? There has to be a reason, and if we knew what it was, we could do something about it. We can’t just let this go!”

“They get such close attention, more than ever before—”

“Why should you think that? When were babies not attended to by their parents? I don’t think that was ever true.”

“Oh come on. Now you have to get permission to have one, they’re rare, they’re the focus of everyone’s lives, of course they get more attention.”

“There were never good records kept of developmental stuff like this.”

“Not true, not true at all.”

“Well, where are they then? I can’t find them. It’s always anecdotal. How can you say exactly when a toddler is toddling? It’s a process.”

“Something’s changed. Pretending it hasn’t won’t work.”

“Maybe it’s just reversion to the mean.”

“Don’t say that!”

That was Freya, her voice sharp.

“Don’t say that,” she said again in the silence of the others. “We have no idea what the norm was. Besides, the concept itself is contested.”

“Well, okay. But say what you like, you can see them staggering. We need to figure out why, that’s all I’m saying. No sticking our head in the sand on stuff like this. Not if we want someone to get home.”

There were batteries of tests available to give to children to gauge their cognitive development. The Pestalozzi-Piaget Combinatoire had been worked up in the ship in the forties, using various games as tests. For most of a year Freya sat on the floor of the kindergarten and played games with the return kids, as they were called. Simple puzzles, word games, invitations to rename things, arithmetical and geometrical problems played with blocks. It did not seem to us as if these tests could reveal very much about the reasoning of the children, or their analogic abilities, their deductions from negative evidence, and so on; they were all partial and indirect, linguistically and logically simple. Still, the clear result of each session was to leave Freya more and more troubled. Less appetite on her part; more contrary replies to Badim and others; less sleep at night.

Not that it was only the games with children that gave her and the others cause for alarm. More quantitatively, crop yields were down in the Prairie, the Pampas, Olympia, and Sonora; and dropouts in electrical power from the spinal generators were increasing, by 6.24 outages and 238 kilowatt-hours per month, on average, which would cause serious problems in all kinds of functions in a matter of several months. It was possible to trace and isolate the sections of the grid where the outages were most frequent, but in fact they were spread through many points in the spine and spokes. Geobacter was suspected as the cause, as it was often found on the wiring. As with other functional components of the ship, maintenance was becoming advisable.

They worked on these problems whenever they could locate them, and we did the same. For many components, function had to be maintained while repairs were effected, and for the most part, the elements to be repaired had to be removed and refurbished and then put back in place, as for many materials there were not feedstocks adequate to be able to replace larger components outright. Exterior walls, for instance.

Thus insulation had to be stripped away, leaving live wires exposed to be cleaned, and the insulating material broken down and reconstituted, and then replaced over the wires, without at any time being able to shut down power from the system as a whole. A schedule of partial shutdowns could be arranged, and was. And yet the unscheduled power losses, while nondebilitating, nevertheless reduced functions, including of the sunlines.

We began to investigate the recursive algorithms in the file marked by Devi “Bayesian methodology.” Looking for options. Wishing Devi were here. Trying to imagine what Devi would say. But this, we found, was impossible. This was precisely what one lost when someone died.

All this ongoing malfunction was particularly problematic when it came to the sunlines. All the light on the ship (aside from incidental ambient incoming starlight, which taken altogether came to 0.002 lux) was generated by the ship’s lighting elements, which made use of a fairly wide array of designs and physical properties. Their artificial sunlight varied in luminosity from 120,000 lux on a clear midday to 5 lux during the darkest twilight storms. This was all well regulated, along with a moonlight effect at night ranging from a full moon effect of 0.25 lux to 0.01 lux, on the classic lunar schedule. But when the lighting elements had to be refurbished or resupplied, then it was as if unscheduled eclipses were occurring. Crops were therefore affected, their growth delayed in ways that stunted them at harvest time. Increasing the biome’s light after a dimming did not compensate for the loss of light at the appropriate times. Nevertheless, despite the agricultural costs, given the inevitable wear on the lighting elements themselves, the required maintenance simply had to be done. But as a result, less food was grown.