"I can do both at once"
The natural defence of drafting is that it costs nothing. You're perfectly capable of holding a reply in mind while following what someone's saying. You've been doing it all your life.
You are doing both, in a sense. But attention is limited, and both tasks draw on the same supply. Levinson and Torreira make this point in the paper I've just described. Understanding speech and producing speech use many of the same mental resources, they note, and so overlapping them can't be free. In the last fraction of a second of someone's turn, the overlap is unavoidable, and it's a small price. When the overlap covers most of what they say, the price goes up.
There's an old and rather neat finding that bears on this. In 1973 Malcolm Brenner published a study of what he called the next-in-line effect. As it's usually described, people sat in a group and took turns to read words aloud, and afterwards they were tested on what they remembered. They recalled the words other people had read reasonably well, with one striking exception. They were poor at recalling the words read out just before their own turn. Charles Bond later ran further experiments on it. I haven't been able to read either paper, so take that as the standard textbook account. The usual interpretation is that in the moments before your own turn, your mind is busy preparing, and what's being said just then doesn't get stored.
If that's right, it explains the phone call in chapter 1 rather exactly. You heard the beginning of what Hannah said perfectly well, the part about the boyfriend and the hot water. That's what set your reply going. What you lost was the stretch immediately before you spoke, because that's when you were holding your opening line ready and listening for her to pause. That stretch contained "everything's a bit much at the moment".
This matters more than it might, because of how people talk. They rarely lead with the important thing. They start with the safe, concrete part of the story, the flatmate and the hot water, and watch how it's received. If it seems to be going all right, they move towards what's really on their mind. So the weightiest sentence in most turns tends to come late. Often it's the last one, and often it's said in a slightly offhand way, as Hannah's was. That's the part of the turn you hear least well when you're drafting.
There's a further, more ordinary cost. Hanne Collins and colleagues, in a set of studies of live conversation that I'll come back to in the next chapter, asked people how much of the time their minds had wandered from what the other person was saying. The average answer was about a quarter of the time. That's the baseline for everybody, before any drafting. If you then give a good share of what's left to rehearsing your reply, you aren't left with much for the person who's talking.
You can test this on the radio or a podcast, where nothing is at stake. Listen to an interview, and whenever you notice an opinion of your own forming about what the guest is saying, pause it and try to repeat the guest's last sentence. Then do the same at a moment when you were simply following. Most people find the second much easier than the first. Nobody was waiting for your reply, and having one in mind still cost you the sentence.
I don't want to overstate this. You're not deaf while you draft. You'd pass a quiz on the gist of what was said. But the gist is what you'd already got by the third sentence, which is why your reply formed then. What drafting costs you is the detail and the late turn, along with the particular word they chose. Often the point of what they were saying was in those.