"I already know what they're going to say"

The second belief is about the guess, and here too there's something right in it.

Stephen Levinson and Francisco Torreira, in 2015, laid out a puzzle about conversation. The gap between turns is about a fifth of a second. But producing even a short utterance takes more than half a second of planning. So the only way people can reply as fast as they do is by predicting where the current turn will end and preparing the reply in advance, while the other person is still talking. Everyone does this. It isn't a habit of interrupters. It's how conversation works at all, and you can't switch it off.

So yes, you know what they're going to say. So does everyone. The prediction is running in every listener at every moment, and it's usually right, in the sense that it gets the shape. You can tell that the sentence is a complaint, or a question, or a story heading for its ending. What the prediction is much worse at is the tail. The last few words. The qualification. The "but" that turns the sentence round. "I got back to the car and there was a note on the windscreen, and I thought it was going to be another one from number twelve, but". You knew the shape. You didn't know the but.

A train announcement is a fair comparison for two sentences. You know from the first three words that it's going to tell you which platform, and that's exactly the part you can't guess.

There's a study of what happens to sentences that get cut off, and it's from medicine, where the stakes make people measure things. Marvel, Epstein, Flowers and Beckman recorded two hundred and sixty-four consultations with family doctors in the United States and Canada, in 1999, and looked at what happened when patients started to say why they'd come. Patients got to finish their opening statement in twenty-eight per cent of the visits. In the rest, the doctor redirected them, after an average of twenty-three seconds. And here is the finding that matters for this chapter. Once redirected, the patients rarely completed what they'd started to say. The tail didn't come later. It was gone. And concerns that surfaced late in the visit, the "oh, and one more thing" at the door, were more common when the doctor hadn't let the patient set out the whole list at the start.

Those were doctors, not dinner tables, and the extension is mine. But the mechanism doesn't seem specific to medicine. A person who is cut off, and then gets the turn back, doesn't go back to where they were. They pick up from where you left them, or they drop it. The tail is the part you couldn't guess, and it's the part that doesn't survive the jump.

Some readers will say at this point that they're simply quicker than most people. Their mind runs ahead, the ending is obvious to them long before it's said, and waiting for it is like waiting for a slow reader to finish a page you've already read. That may be true. It's also beside the point, because the speed of the guess doesn't improve the guess. A fast thinker gets to the shape faster. Nobody gets to the tail faster, because the tail is the part that can't be guessed at any speed. So the quick mind that jumps is quickest at being wrong about the ending, and the quick mind that waits has both the shape and the tail, for the cost of under a second. If you're quick, you're the reader who most needs the pad in chapter 19, because your held thoughts arrive earliest and press hardest.

Go back to the tally from chapter 1. The second column, guesses that turned out right, is your own data on this. If you're like most people, it's shorter than the first column, and even where the guess was right about the shape, you'll notice you don't know what the tail was, because nobody got to say it.