Understanding convictions, your own included

We read a voice more reliably than a face, and we trust it less. The available work suggests this hierarchy is the wrong way round.

Banned for seventy years, confined to fault for a century, settled without a judge since 2017: the history of the exit from marriage throws light on what is expected today of the entrance.

Families do not choose at random between a nursery, a childminder and home care. Any direct comparison first measures that choice.

Absenteeism is measured everywhere, presenteeism almost nowhere. Available estimates nonetheless suggest a cost of the same order.

Contact frequency predicts the duration of a relationship poorly. Available work points to properties less easy to count.

Under the name of psychosocial skills, school curricula have taken up an object that had no place there thirty years ago. The available evaluations are encouraging and limited.

Sharing a home is a test distinct from the feeling, and there is nothing automatic about it. Part of what gets attributed to the couple in fact belongs to domestic organisation.

We regularly overestimate the proportion of people who share our positions. This error of estimation has concrete effects on the way we debate.

The order in which a child sits, crawls and then walks is remarkably constant. The ages at which they do so vary far more than is usually said.

Setting a target changes behaviour in the direction of the target, including when that was not the intention. The mechanism has been documented since the 1970s.

It is not only the number of relationships that differs, it is their reach. A 2022 analysis of massive data put figures on what surveys had suggested.

Fast heart, damp hands, short breath: the body produces roughly the same thing before good and bad news. What differs is what we make of it.

The longer a relationship lasts, the more we believe we can guess what the other is thinking. Confidence in that reading increases; its accuracy does not.

A 2010 article showed that nearly all published work concerned a tiny fraction of humanity. The correction is under way, slowly.

A 2013 study announced that nearly half of American jobs were automatable. Later revisions divided that figure several times over.