Free Embassytown by China Miéville

do’s teach ourselves something with the same noises, which works quite differently. We jury-rigged a methodology, as we had to. Our minds aren’t like theirs. We had to misunderstand Language to learn it.
    When Urich and Becker spoke together with shared, intense feeling, one the Cut and the other the Turn, a flicker of meaning was transmitted, where zettabytes of ’ware had failed.
    Of course they tried again, they and their colleagues practising duets, words that meant hello or we would like to speak . We watched their recorded ghosts. We listened to them learn their lines. “Sounds flawless to me,” said Scile, and even I recognised phrases, but the Ariekei did not. “U and B had no shared mind,” Scile said. “No coherent thoughts behind each word.”
    The Hosts didn’t react with quite the same blankness with which they had heard synthesised voices. They were uninterested in most, but listened hard to a few of these stuttering couples. They didn’t understand it, but they seemed to know that something was being said.
    Linguists, singers, psychospecialists had investigated those pairs who had the most obvious impact. Scientists had striven to work out what they shared. That was how the Stadt Dyadic Empathy Test was created. Attain a certain threshold together on its steep curve of mutual understandingness, fire up machines to connect various brainwaves, synching and linking them, and a particular pair of humans might just be able to persuade the Ariekei that there was meaning to their noises.
    Still communication remained impossible, for megahours after contact. It was a long time after those early revelations that researches into empathy got us anywhere. Very few pairs of people scored well on the Stadt scale, scored highly enough to mum a unified mind behind the Language they ventriloquised. That was the minimum it would take to speak across the species.
    What the colony needed, someone had joked, were single people split in two. And to put it like that was to suggest a solution.
    The first interlocutors with the Hosts were exhaustively trained monozygote twins. Few such siblings could make Language work any better than the rest of us, but those that could were a slightly larger minority than in any control group. They spoke it horribly, we now know, and there were innumerable misunderstandings between them and the Ariekei, but this meant trade, too, at last, and a struggling to learn.
    In my life, I’d met one other pair of idents, non-Embassytowners I mean, in a port on Treony, a cold moon. They were dancers, they did an act. They were blood-born, of course, not made, but still. I was absolutely stunned by them. By how they looked like each other, but only so far. That their hair and clothes were not precisely the same, that they spoke in distinguishable voices, went to different parts of the room, talked to different people.
    On Arieka, for lifetimes, the last two megahours, our representatives hadn’t been twins but doppels, cloned. It was the only viable way. They were bred in twos in the Ambassador-farm, tweaked to accentuate certain psychological qualities. Blood twins had long been outlawed.
    A limited empathy might be taught and drugged and tech-linked in between two people, but that wouldn’t have been enough. The Ambassadors were created and bought up to be one, with unified minds. They had the same genes but much more: it was the minds those carefully nurtured genes made that the Hosts could hear. If you raised them right, taught them to think of themselves right, wired them with links, then they could speak Language, with close enough to one sentience that the Ariekei could understand it.
    The Stadt test was still taken in the out, by students of the psyche and of languages. It had no practical use, now, though—we grew our own Ambassadors in Embassytown, and didn’t have to find each precious potential one among very young twins. As a way to source speakers of Language, the test was