"I know where this is going"

Part of what sets the reply going so early is recognition. Three sentences in, you've placed what you're hearing. It's a flatmate problem, or a difficult boss, or a child who won't sleep. You've heard stories like it before, and you know how they go. Listening to the rest feels like waiting for someone to finish a sentence you could finish for them.

You're often right about the category. It was a flatmate problem. What you're predicting is the general shape of the story, and the general shape isn't what the person rang to tell you.

Nobody phones their parent to convey that flatmates can be inconsiderate. What they want you to know is how this instance is landing on them, and what it's tangled up with in their life. They may also want to tell you what they've already tried, or what they're afraid it says about them. None of that can be predicted from the category. It's in the particulars, and as the last chapter said, the particulars tend to arrive late. When you stop listening at the point of recognition, you've got the part you could have guessed and you've missed everything else.

Finishing the story for them also sends a message, even if you do it silently. When your reply shows that you slotted their situation into a type, they hear that their experience was treated as a standard case. People want to be heard as themselves, and they can tell when they've been treated as an example.

Professionals do the same thing, which should be some comfort. Naykky Singh Ospina and colleagues analysed recordings of 112 medical consultations. In only about a third of them did the clinician ask the patient what they'd come about. When the clinician did ask, the patient was interrupted in two thirds of cases, after a median of eleven seconds. Doctors are trained to recognise patterns, and eleven seconds is often enough to place a complaint in a category. The researchers' concern was that the things which mattered most to patients were getting lost.

One detail in that study is relevant to the fear behind early replies, which is that if you let people talk they'll go on for ever. The patients who weren't interrupted took a median of six seconds to say what they'd come about. Only a small number in the sample were left uninterrupted, so I wouldn't lean on the figure. But it matches a common experience among people who practise listening to the end. Most people, left alone, finish sooner than you'd expect. An interrupted speaker often takes longer overall, because they have to go back and try again to get their point out.

There's a simple way to check your own predictions, and it costs nothing. For the next few conversations, make your guess early about where the speaker is heading, and then listen to the end to see whether you were right. Keep a rough score. My expectation is that you'll get the general drift right most of the time, and that what you miss will be something specific they added, or an emotion, or what they were really asking. You'll also find that the conversations where your guess was wrong were the more important ones.