"I'm fine, I can handle it"

You've been doing these hours for years. You're still standing and you still hit your deadlines. Some people might wilt on fifty-five hours a week, but you seem to be built for it. If the late hours were harming your work, you'd have noticed.

The trouble is that tiredness damages the faculty you'd use to notice. People who are short of rest are poor judges of their own performance. They feel roughly normal while measurably underperforming.

A well-known experiment makes the point with a comparison that's hard to forget. Williamson and Feyer, two Australian researchers, tested thirty-nine people on a set of mental and physical tasks in two conditions. In one, the volunteers were kept awake and tested repeatedly as the hours went by. In the other, the same people were given measured amounts of alcohol and tested at rising blood alcohol levels. That let the researchers express tiredness in units of drink. After seventeen to nineteen hours awake, performance on some tests was as bad as, or worse than, at a blood alcohol level of 0.05 per cent. That's the legal limit for driving in much of Europe. Response speeds on some tests were up to 50 per cent slower.

Do the sum for your own Tuesday. You got up at half past six. When you open the laptop at quarter past nine in the evening, you've been awake for nearly fifteen hours. When you close it at twenty to eleven, you've been awake for just over sixteen. You're approaching the zone that the experiment found comparable to being over the limit. You wouldn't go back to the office after three glasses of wine to finalise the rota. The emails you send at half past ten are written in something like that condition, and you feel perfectly clear-headed as you write them.

A one-off late night does no lasting damage. The concern is what years of it do. Marianna Virtanen and colleagues studied more than two thousand British civil servants in the long-running Whitehall II study. They tested their reasoning, vocabulary and memory, and tested them again five years later. Compared with people working forty hours a week or fewer, those working more than fifty-five hours had lower vocabulary scores at both points, and their scores on the reasoning test declined more over the five years. The difference held after allowing for age, education, occupation, income, health and sleep problems.

This is an observational study, and it can't prove that the hours caused the decline. People who work very long hours may differ in other ways that the researchers couldn't measure. But the finding runs against the flattering idea that long hours keep you sharp.

Errors become more likely too. Dembe and colleagues analysed the job histories of nearly eleven thousand Americans over thirteen years. Jobs with overtime schedules had a 61 per cent higher rate of work injuries and illnesses than jobs without, after allowing for the kind of work involved. Working twelve hours or more a day went with a 37 per cent higher rate. You're unlikely to be injured by a spreadsheet. But an injury at work is mostly a lapse of attention, and the same lapse at a desk produces a wrong figure, a missed clause or a reply sent to the wrong person.

Here's the practical cost, as I see it. The mistakes of the late hours get repaired in the early hours of the next day, which are your best ones. So the evening session doesn't just add little. It borrows against tomorrow, in the form of errors to fix and the dullness of a person who hasn't rested. Over a week, the late hours and the good hours are in competition, and the good hours lose.

There's a simple check. Tomorrow morning, reread the emails you sent after nine tonight, and look again at anything you decided. Most people who do this find they'd have done a fair part of it differently.